AbstractAlthough it is largely accepted that disasters offer windows of opportunity for adoption of risk-reduction measures, the dimensions of support for hazard mitigation following a disaster are not well understood. Drawing on a survey of Upper Texas Coast residents 9 months after Hurricane Harvey (2017), this study explores public support for various types of hazard mitigation within the context of a natural hazard focusing event. Regression models estimate the association of risk perceptions, objective risk, political beliefs, and disaster experience with individual support for hazard mitigation of seven types including structural and nonstructural, household, and community levels. The findings suggest that individuals with higher risk perceptions, those who reside in areas with high flood risk, those who experienced disaster damages to their property, and those who value investment of tax dollars in disaster management are more likely to support hazard mitigation than their counterparts. This support varies across hazard mitigation type, indicating there are some factors related to the benefit gained from the specific mitigation technique that are more relevant than others.IntroductionAlthough disaster management is an area of shared governance among federal, state, and local governments (May and Williams 1986), local governments control some of the most powerful tools of hazard mitigation and are the most proximate level of government connected to publics affected by hazards (Prater and Lindell 2000; Rossi et al. 1982). Yet, due to the complexity of the decentralized policy process, constrained fiscal and human resources, and the tendency for mitigation policies to create winners and losers, social adjustment to hazards is limited (Mileti 1999). Further, there are a number of political challenges that disincentivize local decision makers and citizens from pursuing hazard mitigation, despite the considerable stakes involved with inaction including loss of life and destruction of built and natural environments.Chief among these challenges is the low salience of hazard mitigation on the political agenda. The infrequency of high-impact, low-probability disaster events make reducing hazard risk low on the policy agenda (May and Williams 1986; Prater and Lindell 2000). Hazards typically take a backseat to other concerns perceived as more pressing, such as crime, housing, and education (Burby 2006; Berke and Smith 2009). Second, politicians have little incentive to pursue costly, and often contentious, mitigation policies; rather they are driven to implement strategies that have benefit during their time in office (Mileti 1999; Berke and Smith 2009; Frazier et al. 2013). For example, capital improvement and structural mitigation projects designed to reduce risk typically take years to complete; their benefits can take even longer to realize when the next disaster occurs and are distributed broadly across the community (Prater and Lindell 2000). Therefore, politicians and other groups that bear the political and fiscal costs of the initiative do not directly reap the benefits.Further, federal government incentives have created a moral hazard that undercuts the need for local investment in hazard mitigation. Through federal postdisaster assistance, subsidized flood insurance under the National Flood Insurance Program (NFIP), and homeowner tax credits, there are little incentives for politicians and local residents to spend resources on mitigation (Burby 2006; Berke and Smith 2009). Finally, there is a reluctance among politicians and local decision makers to pursue contentious hazard-mitigation policies such as land-use regulations (Berke and Smith 2009). Despite the fact that land-use and zoning regulations are controlled by local government, there is often a lack of political will to pursue mitigation strategies (Burby 1998; Gerber 2007).The challenges associated with lack of political will to pursue public hazard mitigation—low salience, lack of political incentives, and reluctance to pursue contentious policies—are directly tied to prevailing social attitudes. Studies have shown that local and state politicians align preferences for hazard policy with constituent perceptions (Rossi et al. 1982; Bechtel and Mannino 2020) and that local decision makers see community support as a requirement to pursue hazard mitigation (Frazier et al. 2013). Therefore, lack of concern, awareness, and interest among the public stifles hazard mitigation. However, by the same token, supportive attitudes may bolster public initiatives to reduce risk.It is widely accepted that disasters are focusing events that create windows of opportunity for policy-making by elevating disaster-related issues on the political agenda (Birkland 1997; Kingdon 1995). Policy priorities shift postdisaster as decision makers and the public assess risk and mobilize action (Birkland and Warnement 2017). Although many studies have focused on individual support for household adjustments, the dimensions of support for public hazard mitigation following a disaster are not well understood. To fill this gap, this study examines how risk, political beliefs, and disaster experience are associated with individual support for hazard mitigation of various types.Drawing on a survey of residents in the Houston-Galveston region of Texas 9 months after Hurricane Harvey (2017), the study capitalizes on a natural hazard focusing event where support for mitigation should be strong. The findings suggest that individuals with higher risk perceptions, those who reside in areas with high flood risk, those who have experienced past disaster damages to their property, and those who see the benefit of investment of tax dollars in disaster management are more likely to support hazard mitigation. This support varies across hazard-mitigation technique, indicating that there are factors more relevant in support of different types of risk reduction. These factors largely represent the benefit gained from the specific mitigation technique, which presents an area that future research should further explore.Conceptual Framework of Individual-Level Support for Hazard MitigationRisk perception is the most studied and strongest predictor of household hazard adjustment (Slovic 1987). Protection Motivation Theory (Rogers 1975) has been extensively applied in hazard research to study individual risk perceptions in terms of threat (probability and consequence of the risk) and coping (ability and effectiveness of action to avoid the risk) appraisals (Babcicky and Seebauer 2017; Bubeck et al. 2013; Poussin et al. 2014). Other frameworks, such as the Protection Action Decision Model, have also featured personal risk assessments, and studies have found partial support for the role of risk in the applications of these models (Lindell and Hwang 2008; Lindell and Perry 2012). Research has found that risk perception directly influences (e.g., Grothmann and Reusswig 2006) or mediates the effect of other factors like social capital (e.g., Babcicky and Seebauer 2017), social norms (e.g., Lo 2013), and disaster experience (e.g., Lawrence et al. 2014) on individual decisions for disaster preparedness and risk reduction. However, risk perception alone is insufficient to explain risk reduction, and many empirical studies that have examined risk perceptions and the adoption of household mitigation measures have found weak or no relationship between the two (Bubeck et al. 2012).As an alternative to personal perceptions of risk, some research has examined objective risk as a driver of individual mitigation behavior. It may be that risk perceptions do not fully or accurately capture the risk to which individuals are exposed. For example, individuals may discount or inflate risk in their personal assessments, and people may respond to surveys that measure risk perception in ways that are socially desirable but not necessarily accurate. Early studies of objective risk addressed the role that proximity to streams as well as FEMA designated floodplain plays on individual decisions to purchase flood insurance (Wilcox 1978; Montz 1982). More recent studies have found increasing proportions of floodplain areas are associated with the adoption of hazard insurance (Zahran et al. 2009; Brody et al. 2017a). Additionally, other studies found that household purchases of flood insurance are positively correlated to local mitigation activities while adjusting for objective flood risk indicators like area of floodplains, stream density, and previous property damages from flooding (Zahran et al. 2009).In addition to perceived and objective risk, studies have found that individual adaptation attitudes and behavior can be positively influenced by disaster experience (Cong et al. 2018; Osberghaus 2017). Personal experience with disaster can have a powerful impact on “the willingness to protect oneself from that risk” (Martin et al. 2009, p. 491). In particular, research has shown that people who experience property damages as a result of a disaster event are more likely to undertake risk-reduction actions (Grothmann and Reusswig 2006; Reser et al. 2012; Takao et al. 2004). People who are not affected by a disaster event may underestimate risk and, therefore, be less motivated to support mitigation (Siegrist and Gutscher 2008).Beyond risk and disaster experience, support for hazard mitigation involves politics because it entails policies that determine who gets what, when, and how (Lasswell 1958). Therefore, it is important to consider how individual political beliefs may influence support for risk-reduction activities, ranging from regulations passed by local authorities (e.g., zoning mandates and elevation requirements) to investment of public funds (e.g., construction of levees). In the US, support for government spending is closely tied to political ideology and party identification, with Democrats and liberals generally being more supportive of spending than Republicans and conservatives (McCright et al. 2014). Additionally, recent studies have found that conservatives prefer more proximate levels of government (i.e., state over federal) in managing risk (Choi and Wehde 2019) and tend to view individuals, rather than government, as responsible for disaster preparedness (Wehde and Nowlin 2021). These findings indicate that conservatives have distinct preferences about the locus of responsibility for disaster management, prioritizing lower levels of government and individual responsibility.Based on these studies, we conceptualize support for public hazard mitigation to be associated with risk perception, objective risk, disaster experience, and political attitudes. We thus offer the following hypotheses:H1: Individuals with greater perceived risk are more likely to support public hazard mitigation.H2: Individuals exposed to greater objective risk are more likely to support public hazard mitigation.H3: Individuals who have experienced negative impacts of past disaster events are more likely to support public hazard mitigation.H4: Individuals with conservative political beliefs are less likely to support public hazard mitigation.Data and Measures: Survey of Upper Texas Coast ResidentsStudy SampleTo test the hypotheses outlined, we focus our analysis on three counties located in the Upper Texas Coast: Harris County, Galveston County, and Chambers County (Fig. 1). The Upper Texas Coast is one of the most flood-prone regions in the US, experiencing on average one major hurricane every 15 years (Parisi and Lund 2008). In 2008, Hurricane Ike ravaged the area, making landfall in Bolivar Peninsula with an estimated storm surge of 4.5 m (15 ft), causing approximately $29 billion in losses, and at least 83 fatalities in the state of Texas (Berg 2009; Phan and Airoldi 2015). Most recently, in 2017, Hurricane Harvey made landfall on the southeast Texas coast and moved slowly across the state, leading to widespread flooding that caused over $125 billion in losses (NOAA and NCEI 2021). Given the hazard exposure and experience of this region, the issue of hazard mitigation should be salient among residents and decision makers. However, we recognize that cognitive dissonance processes may occur among populations with high exposure to risk to the point that risk is normalized, discounted, or ignored (Wood and Miller 2020).Data CollectionThe data collected for the study include survey measures of hazard-mitigation support, perceived risk, disaster experience, and political beliefs as well as secondary source information to measure objective risk. The survey employed by the study is an original questionnaire designed to assess coastal hazard mitigation support and willingness to pay for risk reduction, disaster experiences, insurance coverage, and relevant demographic characteristics. The survey was administered to residents of the Upper Texas Coast via phone and online distribution. The survey was in the field by phone May 11 through July 16, 2018, and online August 24 through September 27, 2018. Eligibility requirements included being 18 years or older and a resident of Harris, Galveston, or Chambers Counties. The survey and study protocols were approved by the Texas A&M University Institutional Review Board (#IRB2018-0181D).A total of 2,300 surveys were completed, including 805 phone interviews and 1,495 online surveys. The phone interviews included 142 responses from Chambers County, 251 from Galveston County, and 412 from Harris County. Due to availability of respondents, the online survey was conducted only in Harris and Galveston Counties. A total of 365 online surveys were completed by Galveston County residents, and 1,130 surveys were completed by Harris County residents.The phone survey was conducted by the Public Policy Research Institute at Texas A&M University. The phone sample frame relied on random digit dialing and targeted landlines and cell phones using billing records (to reach residents who live in the area but have numbers associated with other areas). The response rate was 5%, meaning that 5% of all calls to eligible respondents resulted in a completed survey (American Association of Public Opinion, n.d.). Low response rates on phone samples are expected because it is increasingly difficult to contact potential respondents. The phone survey was offered in English and Spanish; a total of 32 surveys were conducted in Spanish. The online survey was limited to English language delivery.The online survey sample was quota-based, drawn from a panel of respondents provided by Qualtrics. Because Qualtrics partners with numerous providers that have proprietary panels across the nation, its panels incorporate participants from online communities, social networks, and websites of all types. The online survey sample matched available Qualtrics panel participants with US Census Bureau data for age, race and ethnicity, and education. Quota-based surveys are increasingly being used to reach participants online. Although participation is improved, the reliance on quota sampling, rather than random sampling, means it is not possible to calculate margins of error for the data that provide a measure of precision. However, nonprobability quota-based surveys offer valid measurements if sample selection and weighting make adjustments that create a representative sample (Pew Research Center 2016).Steps have been taken in this study to adjust the samples to make them representative of the population. A survey weight was constructed for the phone and online samples using population (i.e., county) parameters for age, education, and race and ethnicity available from the US Census Bureau American Community Survey (2016). Then, the phone and online survey weights were combined using accepted techniques (Mercer et al. 2017). The Supplemental Materials provide a detailed description of the survey weight.Dependent Variable Measure: Support for Hazard MitigationMeasures of support for hazard mitigation draw on a set of survey questions that posed the following to respondents: “I’m going to list ways that coastal communities can manage the risk posed by natural hazards. For each one, how much do you support it?” Response options included “do not support,” “support a little,” “support some,” and “support a lot.” This asymmetric scale was purposively designed to maximize measurement of support variability and promote ease of survey task completion while also avoiding validity issues. Recent research has shown that respondents prefer less scale categories, with two- and four-point scales being most preferred (Preston and Colman 2000). Furthermore, studies have found that a neutral (e.g., do not support or oppose) response option may affect the validity of results because respondents may use this for other reasons (i.e., social desirability) than having an intermediate opinion or preference (Garland 1991). Therefore, we deviated from the classic Likert scale as a set of response options in this survey question design.Respondents were asked to rate their support for a number of hazard-mitigation techniques suited to coastal communities: (1) construction of seawalls and levees, (2) creation of retention basins, (3) rehabilitation of natural sand dunes, (4) conservation of wetlands, (5) higher elevation requirements for homes in flood-prone areas, (6) zoning ordinances to guide development, and (7) buyouts of homes flooded multiple times. As indicated in Table 1, these mitigation techniques span two levels—household and regional—as well as two types, namely structural and nonstructural.Table 1. Coastal hazard-mitigation techniquesTable 1. Coastal hazard-mitigation techniquesMitigation techniqueStructuralNonstructuralRegionalHouseholdSeawalls and levees: Seawalls built of concrete such as dikes for preventing wave-based storm surge, and levees for mitigating flooding from streams and rivers. X—X—Basins: Usually in the form of retention or detention ponds which collect, hold or delay the flow of stormwater in flood prone areas.X—X—Sand dunes: Nature-based defense system of dredged sand or a combination of a concrete structure and a sand dune along the coastline.X—X—Wetlands: Protecting and preserving wetlands to help prevent flooding in other areas.— XX—Elevation: Raising buildings on pilings, fill or other support structure to prevent inundation.X—XXZoning: Local planning instrument to reduce flood impacts such as steering development away from physically vulnerable areas.—XX—Buyouts: Acquiring properties that have been repetitively flooded and restoring them to open space. —XXXA seawall is a regional (community)-level structural mitigation measure common in coastal communities. A seawall usually consists of solid or concrete walls that are placed strategically to prevent storm surge or rising waters in the “channel phase” of the flooding (Alexander 1993). Examples of these include dikes, which are usually built to prevent wave-based storm surge, and levees, which prevent flooding from streams and rivers (Brody and Atoba 2018). A similar approach is proposed in our study area called the Ike Dike, a shoreline levee system along Galveston Island and Bolivar Peninsula to reduce the impact of storm surge (Atoba et al. 2018).Basins also protect large areas, especially residential subdivisions from flooding. The most popular forms of basins are retention or detention ponds which collect, hold, or delay the flow of stormwater in flood-prone areas (Lee and Li 2009). These are common in the Houston area, especially for communities where their floodplain has been filled and they are trying to provide compensatory storage to collect excess floodwater.Sand dunes as a mitigation tool also offer protection at the regional level. It is a natural defense system that can be achieved by restoring existing coastline dunes, creating new dunes from dredged sand, or a combination of a concrete structure beneath and a sand dune on top (Brody and Atoba 2018). Wetlands also offer natural defense. Wetlands play a critical role in storing, holding, and slowly releasing floodwaters (Brody et al. 2007). However, the rise in development has led to the destruction of natural wetlands, costing hundreds of thousands of dollars a year in damages (Brody et al. 2011). Protecting existing wetlands or constructing new ones can reverse the adverse impact of development.Elevation of homes is a household mitigation strategy that involves raising properties on pilings, fill, or other support structures to prevent inundation. This additional height is usually expressed as freeboard, which is the number of feet/meters the first floor of a building is raised above the 100-year level of inundation [base flood elevation (BFE)] (Brody and Atoba 2018). Elevating structures can be costly and difficult depending on the existing foundation type. For example, slab-on-grade properties are more expensive to elevate than structures on existing pile or beam foundations (Mobley et al. 2020).Zoning and buyouts are nonstructural mitigation measures. They are also classified as avoidance strategies, which involve directing development away from vulnerable locations (Brody and Atoba 2018). Zoning is usually implemented through local community plans as an instrument to coordinate policies for mitigating flood impacts such as steering development away from physically vulnerable areas. Buyouts occur when local communities give homeowners an opportunity for their properties to be acquired because they have been repetitively flooded. This reduces risk by removing structures and people from highly exposed areas and returning the parcel to natural open space that may absorb the physical impacts of hazards (Tate et al. 2016). Despite this benefit, research has shown that there are limitations to existing buyout policies, making it problematic for homeowners to take part in the process and for local governments to implement (Siders 2019; Atoba et al. 2021).These seven coastal hazard-mitigation techniques offer varying degrees of risk reduction and pose different benefits and costs to residents and decision makers who are responsible for their implementation. Although we recognize these differences, we have no expectations about how the relationship between these and the independent variables examined may vary by mitigation technique. However, we see the value in modeling these separately and learning from the results of these models about the varied dimensions of public support.Independent Variable Measures: Risk, Disaster Experience, and Political BeliefsWe are interested in examining how risk, disaster experience, and political beliefs are associated with support for various hazard-mitigation techniques. Descriptive statistics of all independent variables are given in the Supplemental Materials. To measure risk perception, we rely on the following survey questions: “How likely do you think it is that in the next 10 years there will be a flood that causes…(1) major damage to property in your city? (2) deaths and injuries to people in your community? (3) major damage to your home? (4) disruption to your job that prevents you from working? (5) disruption of electrical, telephone, and other basic services?” Response options included “not at all,” “small extent,” “moderate extent,” “great extent,” and “very great extent.” These questions were replicated from Brody et al.’s (2017b) study of household hazard-mitigation adjustment and focus on flood risk, which is appropriate for the study sample given that all counties included face flood risk, experience floods as part of severe hazard events such as hurricanes, and routinely pursue the mitigation measures detailed previously as ways to reduce flood risk. We combine responses to these questions into a factor score representing latent risk perception. The factor score is highly reliable with a Cronbach’s alpha of 0.834. Factor loadings are given in the Supplemental Materials.In addition to perceived risk, we also account for objective risk as spatially targeted physical exposure to hazards. Exposure “indicates the frequency with which humans in a given area come into contact with hazards” (Allan et al. 2020). Like perceived risk, we focus on flood hazards, adopting the amount of historic flood damages as a proxy for the risk to which individuals are exposed. Historic flood damages are measured as the average of insured flood claim payouts (millions of USD) through the NFIP for the time period 1978 to 2017. It is reported on the ZIP code level.We also considered including the geographical extent of floodplain and percent of open water. Floodplain area was determined using FEMA digitized 100-year floodplain maps of the study area, and percent of open water was calculated from the National Hydrography data set provided by the United States Geological Survey. Although pairwise correlations indicated that all three measures of objective risk may be included in the model, the measures of extent of floodplain and percent open water were not statistically significant in any specification of the model. This may be due to the aggregation of these measures at the ZIP code level. Future research should further consider the most accurate data to measure objective risk or exposure of households in relation to perceived risk.To measure disaster experience, we focus on two recent severe disasters that impacted the entire study area: Hurricane Ike (2008) and Hurricane Harvey (2017). We focus on property damages caused by these events as a proxy of disaster experience. We asked survey respondents: “Thinking back to 2008, what impact did Hurricane Ike have on your home and personal property?” Response options included “no impact,” “positive impact,” “negative impact,” and “not applicable–I did not live in the area then.” We combine responses of “no impact,” “positive impact,” and “not applicable” into one category for purposes of analysis in order to isolate Hurricane Ike damages. We also asked respondents: “On a scale of 0 (none at all) to 100 (extreme devastation), how much did Hurricane Harvey damage your home and personal property?” Given that Hurricane Harvey occurred only a year prior to the survey distribution, we presumed respondents would have better recall and thus allowed for scaled responses to capture the extent of Hurricane Harvey damages.To probe the association of political beliefs with support for hazard mitigation, we include two measures specific to perceptions of the locus of responsibility in disasters that have been shown by recent studies to be associated with conservative political ideology: (1) perceptions of self-reliance, and (2) state and local government responsibility (Choi and Wehde 2019; Wehde and Nowlin 2021). The survey asked respondents: “Do you think individuals should bear the cost of preparing and recovering from natural disasters on their own or tax dollars are well spent on helping communities prepare for and recover from natural disasters.” Responses of “individuals should bear the cost” are considered to measure conservative beliefs of self-reliance. We also asked respondents: “How much responsibility do you think the following have for protecting you from natural hazards…federal government? state government? local government?” Response options included “not at all,” “small extent,” “moderate extent,” “great extent,” and “very great extent.”The measure of state and local responsibility was created by estimating the difference between pairs of federal-state and federal-local responses. The resulting differences were then compared to isolate into one category observations where the respondent holds state and local government more responsible than federal government. The other category included all respondents who believe that state/local government have the same responsibility as federal government or that federal government has more responsibility than the lower levels of government. Focusing on state and local government responsibility is appropriate for this analysis because the hazard-mitigation techniques examined are under the authority of—or require the cooperation of—state and local governments.We also include measures to control for the influence of socioeconomic and demographic characteristics of respondents, including homeowner (compared with renter), racial and ethnic minority group membership (Latino and African American compared with all other races and ethnicities), and college educated (compared with education levels lower than bachelor’s degree). Although past studies have found relationships between these factors and hazard adaptation (Grothmann and Reusswig 2006; Lindell and Hwang 2008), we leave open the expected association of these with support for specific mitigation techniques.Finally, we include county-level fixed effects, measured as the respondent’s county of residence (Harris, Galveston, or Chambers) to control for other factors within those geographic units that may be associated with support for hazard mitigation but are not accounted for in the model. Galveston County and Chambers County are included in the model, with Harris County as the comparison.MethodsAlthough the outcome categories under examination, i.e., “do not support,” “support a little,” “support some,” and “support a lot,” are ordered from least to greatest support, statistical tests indicated that ordinal regression was not an appropriate statistical technique to apply because the parallel regression assumption was violated (Long and Freese 2006). While relaxing this assumption for some variables is an alternative (Fullerton 2009), we decided to treat the dependent variable as a set of discrete choices and estimate a multinominal logistic regression model (McFadden 1974). Multinominal logit models are valid under the assumption of independence of irrelevant alternatives (IIA) that maintains the odds of one outcome do not depend on the other alternatives (Long and Freese 2006). The Hausman test of IAA was estimated, and the results indicated that the assumption was not violated.Varying support for the seven mitigation techniques examined—seawalls and levees, detention basins, sand dunes, wetlands, property elevation, zoning, and home buyouts—was regressed on the following independent variables: risk perceptions, flood damages, Hurricane Ike damage, Hurricane Harvey damage, self-reliance, state and local government responsibility, homeowner, Latino, African American, college educated, Galveston County, and Chambers County. The regression model was specified so that the base category was “do not support,” allowing comparison of opposition to varying levels of support. Correlation and variance inflation factor (VIF) analyses indicated that there was no concern for multicollinearity among the independent variable measures (Supplemental Materials). All regressions estimated apply the survey weight, created to adjust for population parameters on age, education, and race and ethnicity. Note that the number of observations varies across the models due to missing data.The interpretation of the regression results relies on marginal effects. With multinominal logit models, independent variable coefficients can rarely be used to infer the direction of the relationship between an independent variable and the probability of dependent variable outcomes (Bowen and Wiersema 2004; Wulffe 2015). The coefficient of an independent variable represents how the variable relates to the probability of observing a particular dependent variable outcome in relation to the base category. Marginal effects represent the slope of the curve for given predictor values. Therefore, the direction, magnitude, and statistical significance of relationships between independent variables and outcomes are best examined through marginal effects.ResultsExpressed resident support for coastal hazard mitigation varies by strategy, as shown in Fig. 2. Strong support, measured by the response of “support a lot,” is most prevalent for the mitigation technique of higher structural elevation requirements for homes in flood-prone areas (65%), followed by support for conservation of wetlands (54%) and construction of seawalls and levees (50%). Nearing majority support are the mitigation measures of creation of detention basins (47%) and rehabilitation of natural sand dunes (47%). Buyouts of homes flooded multiple times (44%) and zoning ordinances to guide development (38%) garner the lowest strong support and have the highest levels of opposition (10% and 13% “do not support” zoning and home buyouts, respectively). Moderate support, indicated by the response “support some,” is most prevalent for zoning (33%), seawalls and levees (31%), and sand dune rehabilitation (30%). Weak support, measured by the response “support a little,” is also highest for zoning (19%), followed by basins (17%), and home buyouts (16%). What explains the variance in individual support for coastal hazard mitigation? We hypothesize that greater risk (perceived and objective) and disaster experience will be associated with support of mitigation measures. We also expect that conservative political beliefs about disasters, modeled by measures of self-reliance and government responsibility, will be associated with opposition for mitigation.To interpret the results of the multinominal logistic regression model (Supplemental Materials), we turn to marginal effects. Average marginal effects are reported in Table 2 for each dependent variable outcome, i.e., opposition, weak support, moderate support, and strong support, across the seven mitigation strategies examined. Changes for discrete independent variables are reported for the change from the base category, and changes in continuous independent variables are reported for a one-standard-deviation increase from the mean value. Bolded marginal effects indicate statistical significance (p<0.05), and the direction of the association between the independent and dependent variable is represented by sign of the marginal effect.Table 2. Average marginal effectsTable 2. Average marginal effectsIndependent variableDependent variable outcomesm.e.s.e.m.e.s.e.m.e.s.e.m.e.s.e.m.e.s.e.m.e.s.e.m.e.s.e.Perceived riskOpposition−0.021(0.004)−0.021(0.005)−0.024(0.005)−0.023(0.005)−0.021(0.003)−0.036(0.006)−0.040(0.006)Weak support−0.052(0.006)−0.059(0.008)−0.029(0.009)−0.033(0.008)−0.036(0.005)−0.059(0.009)−0.048(0.007)Moderate support−0.052(0.023)−0.049(0.011)−0.049(0.012)−0.024(0.012)−0.052(0.009)−0.025(0.012)−0.038(0.012)Strong support0.124(0.014)0.129(0.014)0.103(0.014)0.079(0.013)0.109(0.011)0.120(0.015)0.126(0.015)Flood damagesOpposition−0.002(0.011)−0.024(0.004)−0.012(0.008)−0.001(0.012)−0.006(0.005)−0.019(0.008)−0.014(0.009)Weak support−0.009(0.013)−0.015(0.011)−0.019(0.011)−0.007(0.010)−0.001(0.010)−0.008(0.013)−0.002(0.012)Moderate support0.007(0.016)−0.008(0.014)−0.003(0.015)−0.009(0.014)0.017(0.013)0.005(0.015)0.002(0.014)Strong support0.004(0.016)0.047(0.015)0.034(0.015)0.018(0.015)−0.010(0.015)0.023(0.016)0.014(0.016)Hurricane Ike damageOpposition−0.004(0.013)−0.012(0.014)−0.002(0.016)−0.029(0.013)0.011(0.013)0.023(0.020)0.010(0.019)Weak support−0.032(0.019)0.039(0.023)−0.023(0.021)0.008(0.020)0.009(0.017)0.025(0.024)−0.004(0.021)Moderate support−0.022(0.028)−0.025(0.027)−0.082(0.026)−0.017(0.026)0.019(0.024)−0.051(0.027)0.026(0.027)Strong support0.058(0.028)−0.001(0.027)0.107(0.027)0.038(0.028)−0.039(0.027)0.003(0.027)−0.032(0.028)Hurricane Harvey damageOpposition0.001(0.006)0.008(0.008)0.008(0.008)0.010(0.008)0.001(0.007)−0.002(0.009)−0.003(0.010)Weak support0.016(0.011)0.010(0.012)0.018(0.011)0.009(0.010)0.016(0.009)0.023(0.013)0.006(0.011)Moderate support0.023(0.014)0.006(0.013)0.035(0.014)0.024(0.014)0.017(0.012)−0.016(0.013)0.017(0.014)Strong support−0.040(0.014)−0.024(0.014)−0.061(0.013)−0.044(0.014)−0.035(0.014)−0.005(0.013)−0.020(0.013)Self-relianceOpposition0.037(0.016)0.043(0.019)0.033(0.021)0.023(0.018)0.041(0.016)0.095(0.025)0.131(0.028)Weak support0.024(0.023)−0.005(0.026)0.019(0.025)0.059(0.026)0.019(0.020)0.016(0.029)0.021(0.025)Moderate support0.010(0.032)0.002(0.031)0.020(0.033)0.010(0.030)0.031(0.028)−0.077(0.030)−0.030(0.030)Strong support−0.071(0.032)−0.040(0.032)−0.072(0.031)−0.092(0.032)−0.091(0.031)−0.033(0.031)−0.122(0.033)State and local government responsibilityOpposition−0.007(0.013)−0.006(0.013)0.005(0.016)−0.015(0.014)−0.022(0.011)0.004(0.018)−0.016(0.018)Weak support−0.040(0.018)0.009(0.023)−0.010(0.021)−0.002(0.020)0.006(0.017)−0.015(0.022)−0.009(0.020)Moderate support0.040(0.028)−0.028(0.026)−0.043(0.026)0.002(0.026)−0.023(0.023)−0.023(0.026)0.013(0.026)Strong support0.006(0.028)0.025(0.028)0.047(0.027)0.015(0.028)0.039(0.027)0.035(0.027)0.013(0.028)HomeownerOpposition−0.049(0.014)−0.020(0.014)−0.055(0.015)−0.045(0.014)−0.035(0.013)−0.061(0.018)−0.036(0.019)Weak support0.010(0.019)−0.045(0.023)−0.004(0.021)−0.035(0.019)−0.015(0.017)−0.008(0.023)0.009(0.021)Moderate support−0.024(0.027)−0.049(0.026)−0.039(0.027)0.007(0.026)0.023(0.023)−0.014(0.027)−0.012(0.026)Strong support0.063(0.028)0.114(0.028)0.098(0.027)0.073(0.027)0.027(0.026)0.083(0.026)0.038(0.027)LatinoOpposition0.027(0.016)0.029(0.016)0.042(0.021)0.046(0.019)0.010(0.013)0.023(0.020)0.019(0.022)Weak support0.051(0.023)0.080(0.026)0.055(0.024)0.054(0.024)0.005(0.019)0.021(0.026)0.040(0.025)Moderate support0.048(0.030)0.010(0.029)0.088(0.030)0.029(0.028)0.032(0.026)0.093(0.030)0.048(0.029)Strong support−0.126(0.031)−0.119(0.031)−0.185(0.029)−0.129(0.030)−0.047(0.030)−0.137(0.029)−0.108(0.030)African AmericanOpposition0.025(0.021)0.035(0.021)0.093(0.030)0.052(0.026)0.007(0.018)0.064(0.028)0.016(0.026)Weak support−0.006(0.026)0.059(0.032)0.019(0.028)0.063(0.029)0.040(0.025)0.045(0.032)0.048(0.030)Moderate support−0.036(0.035)−0.066(0.031)0.000(0.035)0.065(0.035)−0.044(0.028)−0.024(0.035)−0.069(0.031)Strong support0.016(0.036)−0.027(0.034)−0.112(0.033)−0.180(0.034)−0.004(0.035)−0.085(0.032)0.005(0.035)College educatedOpposition−0.008(0.016)−0.039(0.011)−0.048(0.014)−0.021(0.019)0.014(0.017)0.001(0.022)0.005(0.024)Weak support−0.017(0.023)0.014(0.029)−0.024(0.025)−0.046(0.020)−0.027(0.019)0.005(0.029)0.029(0.028)Moderate support0.002(0.033)−0.047(0.031)−0.033(0.033)−0.108(0.028)−0.009(0.028)−0.078(0.031)−0.029(0.030)Strong support0.023(0.035)0.072(0.034)0.104(0.033)0.176(0.033)0.022(0.033)0.073(0.034)−0.005(0.034)Galveston CountyOpposition0.007(0.021)0.078(0.030)−0.002(0.019)0.023(0.027)0.022(0.017)0.027(0.027)0.048(0.027)Weak support−0.030(0.022)−0.007(0.026)−0.044(0.021)−0.021(0.022)−0.013(0.017)0.001(0.029)0.000(0.027)Moderate support−0.143(0.026)−0.007(0.032)−0.045(0.032)−0.060(0.029)−0.049(0.024)0.047(0.034)−0.043(0.028)Strong support0.166(0.033)−0.065(0.033)0.092(0.036)0.058(0.035)0.041(0.031)−0.075(0.030)−0.005(0.033)Chambers CountyOpposition0.047(0.047)0.030(0.048)−0.060(0.013)−0.040(0.024)−0.011(0.024)0.049(0.061)0.036(0.052)Weak support−0.050(0.041)−0.020(0.069)−0.079(0.045)−0.094(0.021)−0.024(0.032)−0.144(0.030)−0.080(0.037)Moderate support−0.089(0.074)−0.132(0.063)−0.160(0.059)−0.011(0.086)−0.095(0.055)0.122(0.095)0.044(0.084)Strong support0.092(0.087)0.122(0.096)0.299(0.064)0.146(0.084)0.129(0.068)−0.027(0.087)0.000(0.076)The average marginal effect of perceived risk is statistically significant across all mitigation measures and for all dependent variable outcomes. Fig. 3 illustrates the estimated marginal effects, surrounded by 95% confidence intervals, at representative values of perceived risk on each dependent variable outcome.Supporting Hypothesis 1, we observe that as perceived risk increases, strong support grows whereas moderate support, weak support, and opposition declines. As shown by the confidence intervals plotted and with the exception of strong support for wetlands and elevation, it is only with increases in perceived risk above the mean value that the predicted probability of strong support is distinguishable from moderate support. Marginal effects indicate that, all else being equal (all other variables held at their means), an individual with perceived risk that is equal to one standard deviation (1.0) greater than the mean (0.0) is 12% more likely to have strong support for seawalls and levees.For detention basins, the difference is 13%; for rehabilitation of sand dunes, 10%; for conservation of wetlands, 8%; for structural elevation, 11%; for zoning, 12%; and home buyouts 13%. These findings align with the research consensus that risk perceptions are one of the most important correlates of protective action in response to a number of climate and natural hazards (e.g., Burnside et al. 2007; Lindell et al. 2009; Lindell and Perry 2012; Whitmarsh 2008). Our findings add nuance to this consensus by showing that high risk perceptions are needed to meaningfully influence support for hazard mitigation.In addition to perceived risk, we also evaluate the association of objective risk with support for hazard mitigation. Recall that historic flood damage in the ZIP code geographic area of the respondent served as our measure of objective risk. The average marginal effects reported in Table 2 indicate that flood damage exhibits a statistically significant association with select dependent variable outcomes for detention basins, sand dune rehabilitation, and zoning. As flood damages increase one standard deviation (3.29 million USD) from the mean (2.75 million USD), the predicted probability of strong support increases by 5% whereas probability of opposition decreases by 2%. For the same change in flood damages, the probability of strong support for sand dune rehabilitation increases by 3%, and the probability of opposition to zoning decreases by 2%. These changes, however, are only a snapshot of the full relationship between flood damages and support/opposition to hazard mitigation.Fig. 4 illustrates the estimated marginal effects, surrounded by 95% confidence intervals, at values representing the full range of flood damages on the dependent variable outcomes. We observe the steepest increase in predicted probability for strong support of basins as flood damages increase from the minimum (0 million USD) to the maximum (12.6 million USD) value plotted. Additionally, the marginal effect graphs reveal that the predicted probability of opposition and weak support of sand dune rehabilitation is not statistically distinguishable at higher values of flood damages, given the overlap in confidence intervals. The predicted probability of moderate support for sand dunes is relatively unchanged across the range of flood damages, whereas strong support increases fairly sharply. For zoning, the predicted probability of moderate and strong support is not statistically distinguishable, whereas moderate support is relatively unchanged across the range of flood damages. Increases in flood damages, however, are associated with statistically significant decreases in the probability of opposition to zoning; an increase across the range of flood damages is associated with a 7% decrease in the probability of opposition. Taken together, these findings offer partial support of Hypothesis 2 and point to the need for additional studies of objective risk and preferences for hazard mitigation.Turning to the association of disaster experience and support for hazard mitigation, marginal effects indicate Hurricane Ike damage is associated with an increase in the predicted probability of strong support for seawalls and levees (6%) and rehabilitation of sand dunes (11%) as well as a decrease in the probability of opposition to conservation of wetlands (3%). In contrast, an increase in the severity of damages from Hurricane Harvey is associated with a decrease in the predicted probability of strong support for seawalls and levees (4%), sand dune rehabilitation (6%), conservation of wetlands (4%), and structural elevation (4%). There is, however, a statistically significant association between increased damages related to Hurricane Harvey and the increased predicted probability of moderate support for sand dune rehabilitation (4%). These results provide partial support for Hypothesis 3.The estimated marginal effects of Hurricane Harvey, surrounded by 95% confidence intervals, across the full range of damage are plotted in Fig. 5. For seawalls and levees, sand dunes, wetlands, and elevation, increased damages are associated with decreased strong support but increased moderate support. These associations have thresholds at which the difference between moderate and strong support is no longer statistically significant: damage values higher than 70 for seawalls and levees, 60 for sand dunes, and 90 for wetlands. Furthermore, the differences in predicted probability of moderate support among damage levels are not statistically significant for the outcome of moderate support for seawalls and levees as well as elevation. For damage values higher than 20, differences in moderate support are also not significant for wetlands. Similarly, differences in predicted probability of opposition across the range of damages are not statistically significant for any of the four mitigation strategies plotted. Above the damage value of 30, weak support for sand dunes is not statistically distinguishable; the same is evident above the damage value of 40 for elevation.The findings related to disaster experience may be largely explained by the nature of the impacts of these two events. Hurricane Ike was a storm surge event that mostly affected residents of Galveston Island. The mitigation measures significantly associated with Hurricane Ike damage are structural mitigation techniques that are implemented directly at the coastline with the intent to reduce hurricane-induced storm surge. Hurricane Harvey was a large 500-year event with unprecedented levels of precipitation that caused widespread damages for residents inland (i.e., the City of Houston) and even outside of frequently flooded communities. The mitigation measures significantly associated with Hurricane Harvey are strategies that may be perceived as better suited for managing coastal (not inland) flooding. The opposition to structural elevation may also reflect divided opinions over regulatory changes that emerged on the local level after Hurricane Harvey to require elevation of structures 0.61 m (2 ft) above the 500-year floodplain (Ross 2019).Admittedly, these event-specific characteristics do not completely explain the differing relationships between Hurricane Ike and Hurricane Harvey disaster experience and support of hazard mitigation. We are, however, confident that the specification of the model is not problematic because we explored the correlation between Hurricane Ike and Hurricane Harvey damages as well as the results of the regressions with alternative specifications of the model, including Hurricane Harvey damage as a categorical variable and exclusion of Hurricane Ike damage. Even with alternative specifications of the model, results remained the same. Therefore, we speculate that measurement may be an issue; certainly, we recognize the measures used in this study are rough approximations of disaster experience. A recent study by Demuth (2018) provided a rich treatment of tornado experience with multiple dimensions, ranging from risk personalization to negative emotional responses. To fully account for the disparate way in which the disaster experience of Hurricane Ike and Hurricane Harvey affects support for (or opposition to) hazard mitigation, a multidimensional measure of disaster experience, such as the one developed by Demuth (2018), should be adapted to flood, hurricane, and storm surge hazards for future data collection efforts and studies.Considering next the association of conservative beliefs and support for hazard mitigation, the regression results indicate that the belief that individuals should be self-reliant by bearing the cost of preparing and recovering from disasters on their own is associated with increased opposition to hazard mitigation. Marginal effects reveal that championing self-reliance over the belief the government, through tax dollars, should cover the costs of disasters is associated with increased probability of opposition (4%) to and decreased probability of strong support (7%) for the construction of seawalls and levees; increased opposition to detention basins (4%); decreased probability of strong support for sand dune rehabilitation (7%); increased probability of weak support (6%) for and decreased probability of strong support (9%) for conservation of wetlands; increased probability of opposition (4%) to and decreased probability of strong support (9%) of structural elevation; increased probability of opposition (10%) to and decreased probability of moderate support (8%) for zoning; and increased probability of opposition (13%) to and decreased probability of strong support (12%) for home buyouts.Although these findings provide robust support for Hypothesis 4, they diverge from recent research. In a study of disaster relief allocation, Bechtel and Mannino (2020) found that affect-based and needs-based norms of distributive justice influenced individual’s preferences for disaster response, whereas political ties did not. The significance of political beliefs in this study may be due to the disaster policies under examination: all are mitigation measures, whereas the Bechtel and Mannino (2020) study focused on disaster relief. Given that mitigation policies are typically implemented under “blue skies,” considerations of affectedness and need in relation to disaster impacts may be less salient whereas political values take precedence. Future work should further explore the political and social values underpinning preferences for disaster policy in varying disaster/emergency management phases.In addition to beliefs about self-reliance, we also examined perceptions of government responsibility. The regression results indicate the belief that state and local government is more responsible than federal government for protecting communities from natural hazards is associated with a decreased probability of weak support (4%) for seawalls and levees and of opposition (2%) to elevation. This may imply that individuals who perceive local and state government as more responsible for disaster management are more likely to support policies under their control. To investigate this further, additional information is needed about the knowledge individuals have pertaining to government authority over and financing of mitigation projects.Turning to our control variables, we observe that homeownership is associated with the increased probability of strong support (6% for seawalls and levees, 11% for basins, 10% for sand dune rehabilitation, 7% for conservation of wetlands, and 8% for zoning) and decreased probability of weak support (5% for basins) or opposition (5% for seawalls and levees, 6% for sand dunes, 5% for wetlands, 4% for elevation, and 6% for zoning) across all of the mitigation strategies examined except for home buyouts. This suggests that homeowners recognize the benefit of risk reduction of a diverse set of mitigation strategies, spanning structural and nonstructural techniques. However, the insignificance of the association of homeownership with home buyouts indicates there is a lack of consensus among homeowners about the value of this mitigation technique.Support for hazard-mitigation measures also varies across demographic, racial, and ethnic lines. College education is associated with a higher probability of strong support for four of the seven mitigation strategies examined (7% retention basins, 10% sand dune rehabilitation, 18% conservation of wetlands, and 7% zoning), and being a racial or ethnic minority is associated with a lower probability of strong support for all mitigation techniques with the exception of structural elevation. Latino is associated with a decreased probability of strong support for the construction of seawalls and levees (13%), detention basins (12%), sand dune rehabilitation (19%), conservation of wetlands (13%), zoning (14%), and home buyouts (11%). Similarly, African American is associated with a decreased probability of moderate support for basins (7%) and home buyouts (7%) as well as strong support for sand dune rehabilitation (11%) and conservation of wetlands (18%).These results may reflect systematic equity issues related to hazards and disasters in the US that privilege the educated and wealthy and disadvantage the poor, uneducated, and racial and ethnic minority groups (Fothergill and Peek 2004). Past research has shown that individuals with the most influence in hazard-mitigation planning are those that can afford the transaction costs of participation, which typically include the well-educated (Godschalk et al. 2003), whereas racial minorities and individuals with lower incomes are less likely to be aware of hazard-mitigation policies and investments (Horney et al. 2015).Furthermore, racial and ethnic minorities are disproportionately affected by natural hazards due to legacies of land use and zoning that have pushed communities of color to settle in areas that are more exposed to hazards (Ross 2019). Despite this heightened exposure, minorities have routinely experienced capital investment diverted from their communities that would reduce hazard risk (Bullard and Wright 2009; Hendricks 2017). Being unsatisfied with government action as well as unaware of and excluded in the planning process may lead racial and ethnic minorities to be less supportive of hazard mitigation.Finally, locational factors also play an important role in support or opposition for flood-mitigation measures. Marginal effects show that being a resident of Galveston County is associated with a 16% increase in the probability of strong support for seawalls and levees and a 6% increase in strong support of sand dune rehabilitation. Residence in Chambers County is associated with a steep increase in the probability of strong support for sand dunes (30%). The higher likelihood of support for these mitigation techniques reflects the benefit these have for coastal communities. Seawalls, levees, and sand dunes tend to provide more direct flood protection to coastal counties than those who are inland. While Galveston County residents support coastline-related mitigation measures, they are less likely to support other risk-reduction techniques such as retention basins and zoning ordinances, strategies that are more often applied further inland.DiscussionThis study has unpacked the broad policy area of hazard mitigation into specific strategies suitable for coastal risk reduction and identified the factors associated with support and opposition of mitigation. The findings show that strong support for structural elevation is by far the highest among the seven strategies evaluated. This is not surprising because it is a mitigation technique that requires no capital investment and affects only those property owners who chose to live in areas exposed to risk. It may be the other mitigation measures evaluated received less support because they involve the use of taxpayer and government funds (e.g., seawalls and levees), are distributive in their benefit (e.g., basin), or involve contentious issues (e.g., zoning).Despite the lower levels of support found for the majority of the mitigation strategies evaluated, the regression results, summarized in Table 3, indicate that hazard mitigation has a base of support. This base of support can be found among individuals who acknowledge and have experienced risk through their risk perceptions, the objective risk of which they are exposed, and the disaster damages they have suffered in the past associated with storm surge events. This base is also robust among homeowners, who value mitigation to protect their property assets. Opposition, on the other hand, is evident among those who hold conservative beliefs, maintaining that individuals should be self-reliant and bear the cost of disasters on their own, as well as among racial and ethnic minorities. Opposition and support, however, are not static across all hazard-mitigation techniques.Table 3. Factors associated with increased likelihood of opposition and strong support for natural hazard–mitigation techniqueTable 3. Factors associated with increased likelihood of opposition and strong support for natural hazard–mitigation techniqueMitigation techniqueIncreased likelihoodPerceivedObjectiveStorm surgeInland floodConservativeHomeownerRacial and ethnic minoritySeawalls and leveesOpposition— ———X——Strong supportX—X——X—BasinsOpposition————X——Strong supportXX———X—Sand dunesOpposition——————XStrong supportXXX——X—WetlandsOpposition——————XStrong supportX————X—ElevationOpposition————X——Strong supportX——————ZoningOpposition————X—XStrong supportX————X—BuyoutsOpposition————X——Strong supportX——————The construction of seawalls and levees as well as the rehabilitation of sand dunes—both structural techniques at the regional level—are supported by individuals who perceive greater risk, those who are exposed to risk as evident by historic flood damages, those who have experienced damages as a result of Hurricane Ike, and homeowners. These individuals are likely to benefit from coastline structural defenses. Similarly, the construction of retention basins and the adoption of zoning regulations to guide development is supported by individuals who have greater risk appraisals and less likely to be opposed by those who reside in a ZIP code area with greater historic flood damages (Table 2). It is likely these individuals appreciate the benefit of these mitigation techniques—that are both on the regional scale—for their own neighborhood or community as a way to reduce risk.Perceived risk was a correlate of strong support for all the hazard-mitigation strategies examined. Given that the data analyzed in this study were collected within a year of a major disaster reinforces that decision makers should seize on windows of opportunity presented by disaster events—when risk perceptions are high—to capitalize on the public’s base of support for hazard mitigation. Yet, we know that many hazard-mitigation policies take considerable time to develop, adopt, fund, and implement. Therefore, it is imperative that social awareness is built about the varied time involved in the mitigation policy process as well as the differences in shared governance required for mitigation measures from start to finish.Of the seven mitigation techniques we evaluated, only two, namely structural elevation and zoning ordinances to guide development, are policies that can be passed and implemented locally without coordination of and funding from state and/or federal agencies. For example, a year after Hurricane Harvey caused extensive damages to the Houston area with unprecedented levels of rainfall and flooding, the City of Houston passed and put into effect regulations that included increased freeboard requirements for homes in the 100-year and 500-year floodplains (Ross 2019). This quick policy response was possible because floodplain regulations are under the control of local authorities, and the City of Houston negotiated to manage federal disaster relief assistance directly, rather than through the state.On the other hand, large-scale regional structural mitigation projects, including seawalls and levees, can take decades to develop and require considerable interagency, often spanning multiple levels of government, collaboration. In Texas, the Ike Dike concept—featuring a coastal spine and gate system to protect against storm surge—was conceived in response to the extensive damage and death caused by Hurricane Ike in 2008. Ten years later the US Army Corps of Engineers, in conjunction with the Texas General Land Office, announced a plan to move the Ike Dike project forward along with ecosystem restoration and beach and dune nourishment (Powell 2018). A revised plan was put to the public in 2020, and funding mechanisms have been proposed by the state legislature in 2021 that would provide the administrative structures needed for the $26 billion project (Powell 2021).Building awareness of how mitigation techniques vary in terms of which level of government has authority and capacity for that specific policy is challenging because individuals may not have a sense of “intuitive federalism” or discernment about government actions at the level in which that action is likely to take place (Wehde and Nowlin 2021). Social awareness about the management of hazard risk can be cultivated through public participation in the planning process to involve not only information sharing but, importantly, consensus-building processes that provide citizens the opportunity to have ownership in hazard-mitigation policy (Pearce 2003). Given our finding that racial and ethnic minorities oppose or are less likely to support hazard mitigation (Table 2), it is imperative that often-left-out groups are included and engaged in participatory planning processes.Although such participatory planning offers considerable potential for shared governance of risk, there is a need in Texas, as well as other places, to build capacity for locally created and externally funded mitigation strategies that are forward-thinking and fill in the gaps of state systems (Villegas et al. 2018). This requires the integration of comprehensive planning and community development with hazard-mitigation planning in a way that affords a space for public involvement and enables local government action. Localities that are allowed to take a flexible approach to hazard mitigation, rather than a project-oriented approach, have been found to adopt a wider range of policies to address risk reduction (Lyles et al. 2014). This points to the need to consider changes to state approaches to local hazard-mitigation planning mandates.Changes to the financing of disaster management is another alternative that may involve replacing the existing cost-share system with a sliding scale to account for true risk or creating regional disaster organizations that allow localities to take a larger role in managing regional recovery and mitigation (Villegas et al. 2018). If such institutional changes are successful in incentivizing local mitigation action, they may also shift the dimensions of individual support for different hazard-mitigation techniques as calculations of responsibility attribution change and become more salient on the local level.Although our study offers insights into the nuances of hazard-mitigation support, it is subject to a few limitations. The data employed were collected within a year of a major disaster event (Hurricane Harvey). We acknowledge that disaster issues were highly salient during this time period and that risk perceptions may have been higher than usual. Although we see this as a snapshot in time ripe for capturing public perceptions in that window of opportunity presented by disasters, we recognize that this limits the generalizability of the findings.Likewise, the data are limited in generalizability with a focus on three counties in the Upper Texas Coast. Although flood-prone and exposed to considerable hazard risk, we acknowledge that residents of Harris, Galveston, and Chambers Counties of Texas are not representative of the broad US population and that the findings must be understood as indicative of this region’s perceptions of risk and preferences for policy. Research has shown that residents of the Houston/Galveston area tend to perceive more flood risk than exposure suggests, which may indicate a unique disaster subculture in this area (Allan et al. 2020).The data are also limited by the underrepresentation of Latinos in the survey sample. Although this was addressed in the analysis through the survey weight, the sampling approach could have been improved to encourage more Latino participants. Namely, offering the online survey in English and Spanish, in addition to the dual-language phone survey, might have improved the participation of Latino respondents. Another limitation to the data is its cross-sectional structure. Cross-sectional survey responses cannot establish causality and may be biased in self-reporting (e.g., social desirability). Furthermore, there are potential problems associated with using political boundaries such as ZIP codes to measure objective risk. However, ZIP-code-level spatial aggregations provide valuable insight into household-level vulnerabilities and have also been used in similar studies to measure objective flood risk at the household level such as floodplain area and proximity to open waters (Harlan et al. 2019; Botzen et al. 2009). To address these concerns, future studies should focus on capturing risk perceptions across time and measuring objective flood risk at the hydrological unit such as the watershed scale.ConclusionMany of the challenges local hazard-mitigation policy face are sociopolitical: low policy salience, lack of political incentives, and reluctance to pursue contentious policies. Despite these and counter to the thinking that this makes hazard mitigation a “policy without a public” (May 1991), this study provides evidence that there is support for hazard-mitigation policy. Using a survey of over 2,000 respondents from the Upper Texas Coast and secondary data to measure objective risk on the ZIP code level, we find that individuals with greater risk perception, greater exposure to flood risk, experience with disaster damages from a storm surge event, and homeowners are more likely to offer strong support for hazard mitigation. Individuals who have experienced greater damages from an inland flooding event, hold conservative beliefs of self-reliance in disasters, and are racial and ethnic minority group members are more likely to offer weak or no support for mitigation measures to address coastal hazards.Seven hazard-mitigation strategies are evaluated in this study: construction of seawalls and levees, creation of retention basins, rehabilitation of natural sand dunes, conservation of wetlands, higher elevation requirements for homes in flood-prone areas, zoning ordinances to guide development, and buyouts of homes flooded multiple times. Support for specific policies seems to align with the benefit ascribed to the measure. For example, individuals that experienced property damages from Hurricane Ike are more likely to support seawalls and levees as well as sand dunes—all coastline, regional measures that would reduce the risk of storm surge events like Hurricane Ike.The findings of this analysis are ripe for future research leveraging data beyond the scope of this study to explore how individuals appraise the benefit and cost of, as well as norms of, distributive justice (e.g., affectedness-based or needs-based) (Bechtel and Mannino 2020) associated with specific mitigation measures. Such research will only become more important as climate change increases the frequency and intensity of natural hazards communities face, thereby pressuring governments to make hazard mitigation a policy priority.Data Availability StatementData, models, or code for the analyses presented in this study are available from the corresponding author upon reasonable request.AcknowledgmentsThe authors would like to acknowledge funding support of the Texas General Land Office. All findings, conclusions, or recommendations expressed in this research are those of the authors and do not necessarily reflect the view of the funding agency. The authors also acknowledge Abbey Hotard, doctoral student at Texas A&M University at Galveston, for creation of the study site map.References Alexander, D. 1993. Natural disasters. London: Routledge. Allan, J. N., J. T. Ripberger, W. Wehde, M. Krocak, C. L. Silva, and H. C. Jenkins-Smith. 2020. “Geographic distributions of extreme weather risk perceptions in the United States.” Risk Anal. 40 (12): 2498–2508. https://doi.org/10.1111/risa.13569. Atoba, K. O., S. D. Brody, W. E. Highfield, and W. J. Merrell. 2018. “Estimating residential property loss reduction from a proposed coastal barrier system in the Houston-Galveston region.” Nat. Hazards Rev. 19 (3): 05018006. https://doi.org/10.1061/(ASCE)NH.1527-6996.0000297. Atoba, K. O., G. Newman, S. D. Brody, W. E. Highfield, Y. Kim, and A. Juan. 2021. “Buy them out before they are built: Evaluating the proactive acquisition of vacant land in flood-prone areas.” Environ. Conserv. 2021 (1): 1–9. https://doi.org/10.1017/S0376892921000059. Babcicky, P., and S. Seebauer. 2017. “The two faces of social capital in private flood mitigation: Opposing effects on risk perception, self-efficacy and coping capacity.” J. Risk Res. 20 (8): 1017–1037. https://doi.org/10.1080/13669877.2016.1147489. Bechtel, M., and M. Mannino. 2020. “Retrospection, fairness, and economic shocks: How do voters judge policy responses to natural disasters?” Political Sci. Res. Methods 2017 (Apr): 1–19. https://doi.org/10.1017/psrm.2020.39. Berke, P., and G. Smith. 2009. “Hazard mitigation, planning, and disaster resiliency: Challenges and strategic choices for the 21st century.” In Building safer communities: Risk governance, spatial planning and responses to natural hazards, edited by U. F. Paleo, 1–18. Amsterdam, Netherlands: IOS Press. Birkland, T. A. 1997. After disaster: Agenda setting, public policy and focusing events. Washington, DC: Georgetown University Press. Birkland, T. A., and M. K. Warnement. 2017. “Focusing events, risk, and regulation.” In Policy shock: Recalibrating risk and regulation after oil spills, nuclear accidents and financial crises, edited by E. J. Ballesisen, L. S. Bennear, K. D. Krawiec, and J. B. Wiener, 107–128. Cambridge, UK: Cambridge University Press. Botzen, W. J. W., J. C. J. H. Aerts, and J. C. J. M. Van Den Bergh. 2009. “Dependence of flood risk perceptions on socioeconomic and objective risk factors.” Water Resour. Res. 45 (Oct): 10. https://doi.org/10.1029/2009WR007743. Bowen, H. P., and M. F. Wiersema. 2004. “Modeling limited dependent variables: Methods and guidelines for researchers in strategic management.” In Vol. 1 of Research methodology in strategy and management, edited by D. J. Ketchen and D. D. Bergh, 87–134. Bingley, UK: Emerald Group. Brody, S., K. Atoba, A. Sebastian, W. Highfield, R. Blessing, W. Mobley, and L. Stearns. 2021. “A comprehensive framework for coastal flood risk reduction: Charting a course towards resiliency.” In A blueprint for coastal adaptation, edited by C. Kousky, B. Fleming, and A. Berger, 2–28. Washington, DC: Island Press. Brody, S. D., and K. O. Atoba. 2018. “Institutional resilience: The example of flood resiliency in the United States.” In Vulnerability and resilience to natural hazards, edited by S. Fuchs and T. Thaler, 237–256. Cambridge, UK: Cambridge University Press. Brody, S. D., W. E. Highfield, and J. E. Kang. 2011. Rising waters: The causes and consequences of flooding in the United States. Cambridge, UK: Cambridge University Press. Brody, S. D., W. E. Highfield, H. C. Ryu, and L. Spanel-Weber. 2007. “Examining the relationship between wetland alteration and watershed flooding in Texas and Florida.” Nat. Hazard. 40 (2): 413–428. https://doi.org/10.1007/s11069-006-9003-3. Brody, S. D., W. E. Highfield, W. Wilson, M. K. Lindell, and R. Blessing. 2017a. “Understanding the motivations of coastal residents to voluntarily purchase federal flood insurance.” J. Risk Res. 20 (6): 760–775. https://doi.org/10.1080/13669877.2015.1119179. Brody, S. D., Y. Lee, and W. E. Highfield. 2017b. “Household adjustment to flood risk: A survey of coastal residents in Texas and Florida, United States.” Disasters 41 (3): 566–586. https://doi.org/10.1111/disa.12216. Bubeck, P., W. J. Botzen, H. Kreibich, and J. C. Aerts. 2013. “Detailed insights into the influence of flood-coping appraisals on mitigation behaviour.” Global Environ. Change 23 (5): 1327–1338. https://doi.org/10.1016/j.gloenvcha.2013.05.009. Bullard, R. D., and B. Wright. 2009. Race, place, and environmental justice after Hurricane Katrina: Struggles to reclaim, rebuild, and revitalize New Orleans and the Gulf Coast. New York: Perseus Books. Burby, R. J. 1998. “Natural hazards and land use: An introduction.” In Cooperating with nature: Confronting natural hazards with land use planning for sustainable communities, edited by R. J. Burby, 1–26. Washington, DC: Joseph Henry Press. Burby, R. J. 2006. “Hurricane Katrina and the paradoxes of government disaster policy: Bringing about wise governmental decisions for hazardous areas.” Ann. Am. Acad. Political Soc. Sci. 604 (1): 171–191. https://doi.org/10.1177/0002716205284676. Burnside, R., D. S. Miller, and J. D. Rivera. 2007. “The impact of information and risk perception on the hurricane evacuation decision-making of greater New Orleans residents.” Sociol. Spectrum 27 (6): 727–740. https://doi.org/10.1080/02732170701534226. Choi, J., and W. Wehde. 2019. “Venue preference and earthquake mitigation policy: Expanding the micro-model of policy choice.” Rev. Policy Res. 36 (5): 683–701. https://doi.org/10.1111/ropr.12354. Cong, Z., A. Nejat, D. Liang, Y. Pei, and R. J. Javid. 2018. “Individual relocation decisions after tornadoes: A multi-level analysis.” Disasters 42 (2): 233–250. https://doi.org/10.1111/disa.12241. Demuth, J. L. 2018. “Explicating experience: Development of a valid scale of past hazard experience for tornadoes.” Risk Anal. 38 (9): 1921–1943. https://doi.org/10.1111/risa.12983. Garland, R. 1991. “The mid-point on a rating scale: Is it desirable?” Marketing Bull. 2 (Jan): 66–70. Godschalk, D. R., S. Brody, and R. Burby. 2003. “Public participation in natural hazard mitigation policy formation: Challenges for comprehensive planning.” J. Environ. Plann. Manage. 46 (5): 733–754. https://doi.org/10.1080/0964056032000138463. Grothmann, T., and F. Reusswig. 2006. “People at risk of flooding: Why some residents take precautionary action while others do not.” Nat. Hazards 38 (1–2): 101–120. https://doi.org/10.1007/s11069-005-8604-6. Harlan, S. L., M. J. Sarango, E. A. Mack, and T. A. Stephens. 2019. “A survey-based assessment of perceived flood risk in urban areas of the United States.” Anthropocene 28 (2): 100217. https://doi.org/10.1016/j.ancene.2019.100217. Horney, J., M. Simon, S. Grabich, and P. Berke. 2015. “Measuring participation by socially vulnerable groups in hazard mitigation planning, Bertie County, North Carolina.” J. Environ. Plann. Manage. 58 (5): 802–818. https://doi.org/10.1080/09640568.2014.892870. Kingdon, J. W. 1995. Agenda, alternatives and public policies. 2nd ed. New York: Harper Collins. Lasswell, H. D. 1958. Politics: Who gets what, when, how. New York: Meridian Books. Lawrence, J., D. Quade, and J. Becker. 2014. “Integrating the effects of flood experience on risk perception with responses to changing climate risk.” Nat. Hazards 74 (3): 1773–1794. https://doi.org/10.1007/s11069-014-1288-z. Lee, J. S., and M. Li. 2009. “The impact of detention basin design on residential property value: Case studies using GIS in the hedonic price modeling.” Landscape Urban Plann. 89 (1): 7–16. https://doi.org/10.1016/j.landurbplan.2008.09.002. Lindell, M. K., S. Arlikatti, and C. S. Prater. 2009. “Why people do what they do to protect against earthquake risk: Perceptions of hazard adjustment attributes.” Risk Anal. 29 (8): 1072–1088. https://doi.org/10.1111/j.1539-6924.2009.01243.x. Long, J. S., and J. Freese. 2006. Regression models for categorical dependent variables using Stata. College Station, TX: Stata Press. Martin, W. E., I. M. Martin, and B. Kent. 2009. “The role of risk perceptions in the risk mitigation process: The case of wildfire in high risk communities.” J. Environ. Manage. 91 (2): 489–498. https://doi.org/10.1016/j.jenvman.2009.09.007. May, P. J. 1991. “Addressing public risks: Federal earthquake policy design.” J. Policy Anal. Manage. 10 (2): 263–285. https://doi.org/10.2307/3325175. May, P. J., and W. Williams. 1986. Disaster policy implementation: Management strategies under shared governance. New York: Plenum Press. McCright, A. M., C. Xiao, and R. E. Dunlap. 2014. “Political polarization on support for government spending on environmental protection in the USA, 1974–2012.” Soc. Sci. Res. 48 (Nov): 251–260. https://doi.org/10.1016/j.ssresearch.2014.06.008. McFadden, D. 1974. “Conditional logit analysis of qualitative choice behavior.” In Frontiers of econometrics, edited by P. Zarembka. New York: Academic Press. Mercer, A. W., K. Frauke, K. Scott, and E. A. Stuart. 2017. “Theory and practice in nonprobability surveys: Parallels between causal inference and survey inference.” Public Opin. Q. 81 (1): 250–271. https://doi.org/10.1093/poq/nfw060. Mileti, D. S. 1999. Disasters by design: A reassessment of natural hazards in the United States. Washington, DC: Joseph Henry Press. Mobley, W., K. O. Atoba, and W. E. Highfield. 2020. “Uncertainty in flood mitigation practices: Assessing the economic benefits of property acquisition and elevation in flood prone communities.” Sustainability 12 (5): 2098. https://doi.org/10.3390/su12052098. NOAA and NCEI (National Oceanic and Atmospheric Administration and National Centers for Environmental Information). 2021. “US billion-dollar weather and climate disasters.” Accessed December 10, 2021. https://www.ncdc.noaa.gov/billions/. Pearce, L. 2003. “Disaster management and community planning, and public participation: How to achieve sustainable hazard mitigation.” Nat. Hazards 28 (2): 211–228. https://doi.org/10.1023/A:1022917721797. Preston, C. C., and A. M. Colman. 2000. “Optimal number of response categories in rating scales: Reliability, validity, discriminating power, and respondent preferences.” Acta Psychol. 104 (1): 1–15. https://doi.org/10.1016/S0001-6918(99)00050-5. Reser, J. P., G. L. Bradley, M. C. Ellul, and R. Callaghan. 2012. Public risk perceptions, understandings and responses to climate change and natural disasters in Australia and Great Britain. Gold Coast, Australia: National Climate Change Adaptation Research Facility. Ross, A. D. 2019. “Hurricane Harvey: Issues for urban development.” In US emergency management in the 21st century: From disaster to catastrophe, edited by C. B. Rubin and S. L. Cutter. New York: Routledge. Rossi, P. H., J. D. Wright, and W.-B. Eleanor. 1982. Natural hazards and public choice: The state and local politics of hazard mitigation. New York: Academic Press. Slovic, P. 1987. “Perception of risk.” Science 236 (4799): 280–285. Takao, K., T. Motoyoshi, T. Sato, and T. Fukuzono. 2004. “Factors determining residents’ preparedness for floods in modern megalopolises: The case of the Tokai flood disaster in Japan.” J. Risk Res. 7 (7–8): 775–787. https://doi.org/10.1080/1366987031000075996. Wehde, W., and M. C. Nowlin. 2021. “Public attribution of responsibility for disaster preparedness across three levels of government and the public: Lessons from a survey of residents of the US South Atlantic and Gulf Coast.” Publius: J. Federalism 51 (2): 212–237. https://doi.org/10.1093/publius/pjaa037. Whitmarsh, L. 2008. “Are flood victims more concerned about climate change than other people? The role of direct experience in risk perception and behavioural response.” J. Risk Res. 11 (3): 351–374. https://doi.org/10.1080/13669870701552235. Wilcox, W. H. 1978. “FDAA prospective—An invited comment.” Nat. Hazard. Observer 2 (4): 1–2. Zahran, S., S. Weiler, S. D. Brody, M. K. Lindell, and W. E. Highfield. 2009. “Modeling national flood insurance policy holding at the county scale in Florida, 1999–2005.” Ecol. Econ. 68 (10): 2627–2636. https://doi.org/10.1016/j.ecolecon.2009.04.021.