PUBLIC OPINION ON CLIMATE CHANGE

Existing research identifies three major correlates of climate attitudes: demographics, risk perceptions, and personal experiences (Egan and Mullin Reference Egan and Mullin2017; Hornsey et al. Reference Hornsey, Harris, Bain and Fielding2016). In addition to shaping individual perceptions about climate mitigation, these factors also affect whether and how individuals attribute extreme weather to the effects of climate change (Ogunbode et al. Reference Ogunbode, Demski, Capstick and Sposato2019). Among the U.S. public, partisanship, education, and gender are the strongest demographic predictors of climate beliefs. The effect of partisanship is unsurprising given polarization of elite rhetoric and trust in science (McCright and Dunlap Reference McCright and Dunlap2011). Numerous studies show liberals are more supportive of climate change mitigation (Boudet et al. Reference Boudet, Giordono, Zanocco, Satein and Whitley2020; Hazlett and Mildenberger Reference Hazlett and Mildenberger2020). Similarly, women (Bush and Clayton Reference Bush and Clayton2023; Leiserowitz Reference Leiserowitz2006) and better-educated individuals (Hornsey et al. Reference Hornsey, Harris, Bain and Fielding2016) are more likely to believe in climate change and support mitigation.

Beyond demographics, risk perceptions—subjective judgments of threat—also influence climate attitudes. Given the long-standing, diffuse threat posed by climate change, values and worldviews powerfully shape climate risk perceptions (Leiserowitz Reference Leiserowitz2006). For instance, people who value hierarchical social organization are less supportive of climate mitigation (Hornsey et al. Reference Hornsey, Harris, Bain and Fielding2016). Likewise, empaths are more supportive of addressing climate displacement (Arias and Blair Reference Arias and Blair2022). Beyond affect, risk perceptions are also shaped by geographic vulnerability and personal experiences. In particular, experiences of extreme weather are associated with increasing belief in climate change (Bergquist and Warshaw Reference Bergquist and Warshaw2019; Brody et al. Reference Brody, Zahran, Vedlitz and Grover2008; Howe et al. Reference Howe, Marlon, Mildenberger and Shield2019; Sambrook et al. Reference Sambrook, Konstantinidis, Russell and Okan2021). In this article, we focus on the attitudinal consequences of Hurricane Ian.

CLIMATE DISASTERS AND CLIMATE MIGRATION

While much academic and policy attention is paid to public opinion on migration (Hainmueller and Hopkins Reference Hainmueller and Hopkins2014) and climate change generally (Egan and Mullin Reference Egan and Mullin2017), little work considers public opinion on climate displacement.Footnote ⁹ To be sure, interdisciplinary scholars have recognized important dynamics related to climate migration. Lawyers have theorized how climate migrants could be integrated into migration conventions (McAdam Reference McAdam2012), and political theorists have weighed moral obligations states have vis-à-vis the climate-displaced (Draper Reference Draper2022). Likewise, economists and demographers have studied the effects of climate change on migration (Hunter, Luna, and Norton Reference Hunter, Luna and Norton2015), and conflict scholars have examined tensions between climate migrants and hosts (Bhavnani and Lacina Reference Bhavnani and Lacina2015; Koubi et al. Reference Koubi, Böhmelt, Spilker and Schaffer2018).Footnote ¹⁰ However, systematic analyses of public opinion on climate-induced migration are rare. In particular, we offer—to the best of our knowledge—the first study on how climate disasters affect attitudes on climate migration. This is crucial because environmental disasters are the leading cause of climate-related displacement globally and because public opinion on climate migrants is central to understanding the prospects for their integration in receiving communities (Obokata, Veronis, and McLeman Reference Obokata, Veronis and McLeman2014).

THEORY

We draw on this latter account and wed it with insights from aforementioned scholarship on the attitudinal effects of disaster exposure to understand how climate disasters shape climate migration attitudes, in addition to general climate beliefs.Footnote ¹³ We specifically consider exposure and attitudes in the context of Hurricane Ian. Though prior studies have examined a variety of climatic events, relatively little work considers hurricanes (but see Bergquist, Nilsson, and Wesley Schultz Reference Bergquist, Nilsson and Wesley Schultz2019; Sloggy et al. Reference Sloggy, Suter, Rad, Manning and Goemans2021). This is important because individuals respond differently to different types of climatic phenomena (Howe et al. Reference Howe, Marlon, Mildenberger and Shield2019). Moreover, hurricanes are the leading cause of internal climate displacement in the United States.

We argue that personal experiences with climate disasters like hurricanes sharpen risk perceptions and make climate change’s impacts more concrete. While climate migration and climate change are conceptually abstract, hurricanes are tangible. Because hurricanes create substantial migratory pressures, they are particularly likely to spur mass evaluations of climate displacement. People in the path of the storm must weigh the costs and risks of fleeing versus remaining, while those in storm-adjacent regions must consider how their communities will respond to potential local influxes of climate-displaced individuals (Hopkins Reference Hopkins2012).

Empathy undergirds favorability toward climate migrants (Arias and Blair Reference Arias and Blair2022), and personal exposure to displacement-inducing storms is particularly likely to stimulate empathic perspective-taking. In much the way that displacement experiences mobilize pro-social refugee attitudes (Hartman and Morse Reference Hartman and Morse2020; Williamson et al. Reference Williamson, Adida, Lo, Platas, Prather and Werfel2021), we expect hurricane victims to reflect on their disaster experiences and become more supportive of policies to benefit climate migrants. Importantly, a mobilizing effect of hurricanes on climate migration attitudes is also consistent with rational retrospection. After disasters, victims in climate-affected regions should be more conscious of future climate displacement-related risks and hence more supportive of ameliorative policies.Footnote ¹⁴

While Americans’ broader climate opinion is sticky (McCright and Dunlap Reference McCright and Dunlap2011), we argue that personal experiences with hurricanes should also shock these attitudes, increasing support for climate mitigation and belief in anthropogenic climate change. Hurricanes represent a salient manifestation of the risks posed by unmitigated global warming. The severe damage they cause should concretize the high relative costs of climate change for individuals exposed. Rational voters in climate-affected communities should also reward mitigatory policies that reduce the effects of climate change (Gasper and Reeves Reference Gasper and Reeves2011) and especially the risks of future hurricanes.

While we did not preregister hypotheses about the durability of these effects, literature on recency bias (e.g., Arndt, Jensen, and Wenzelburger Reference Arndt, Jensen and Wenzelburger2021; Nordhaus Reference Nordhaus1975) suggests attitudinal consequences of Hurricane Ian are likely to decay quickly. We offer exploratory evidence on this question.

A QUASI-EXPERIMENT ON HURRICANE EXPOSURE

To test our theory, we administered a preregistered survey on Lucid—a well-known, online platform (Coppock and McClellan Reference Coppock and McClellan2019)—during 2022’s Atlantic hurricane season.Footnote ¹⁵ Our survey targeted four states most vulnerable to hurricanes: Florida, Louisiana, Texas, and North Carolina.Footnote ¹⁶ We used quota sampling to obtain a respondent pool from these states that approximates the adult population of the United States with respect to census benchmarks for race, ethnicity, gender, and age (Table SI-1).Footnote ¹⁷ Given this design, effects may not generalize to the U.S. population as a whole. Still, this particular sample is interesting and important: individuals in the focal states are swing voters cross-pressured by climate change and migration. This makes ours a hard case in which to detect positive effects of hurricane exposure on climate opinion.

From August 11 to October 28, 2022, we fielded a weekly, cross-sectional survey ( $ \approx $ 250 respondents) across these states.Footnote ¹⁸ This time frame represents the historically most active period of hurricane season, and we (correctly) anticipated that our survey would be interrupted by a storm. Over the study’s duration, we captured 3,202 respondents geolocated to the four states of interest. Following our preregistration plan, we filtered out respondents who finished in the top and bottom deciles of survey duration or who reported ages less than 18 or greater than 99. This left a final sample of 2,563 respondents. Figure 1 maps the distribution of these respondents.

Figure 1. Geographic Distribution of Survey Respondents

Note: The dashed red line marks Hurricane Ian’s eyepath.

We also fielded an exploratory, follow-up survey from March 7–10, 2023. This follow-up was administered using the same specifications as the original survey and was designed to measure the durability of Hurricane Ian’s impacts. In the follow-up wave, we captured 847 new respondents geolocated to the focal states. Applying the same filters on duration and age from the main sample gave a final follow-up sample of 715 respondents (Figure A-12).

In our surveys, respondents were first asked a demographic battery to gather information on relevant traits and beliefs. Subsequently, outcome variables were measured, with question order randomized across respondents. Section SI-9 reports the questionnaire. We study 27 items (Table 1) related to seven focal concepts of interest: (1) issue importance of climate migration (Cronbach’s $ \alpha $ = 0.57); (2) policy action on climate migration ( $ \alpha $ = 0.74); (3) issue importance of climate change ( $ \alpha $ = 0.72); (4) policy action on climate change ( $ \alpha $ = 0.76); (5) climate mitigation policies ( $ \alpha $ = 0.90); (6) climate adaptation policies ( $ \alpha $ = 0.90); and (7) belief in climate science ( $ \alpha $ = 0.70). For each concept, we asked three to five questions designed to elicit theoretically relevant opinions.Footnote ¹⁹ As specified in our pre-analysis plan, we combined related items into summary indices, which organize our findings and reduce multiple inference concerns. Each index is the average of standardized outcomes weighted by the inverse covariance matrix (Anderson Reference Anderson2008).Footnote ²⁰ Cronbach’s $ \alpha $ (reported above) and principal component analyses (Table A-5) confirm our indices are reliable and unidimensional.

Table 1. Coding of Dependent Variables

Note: We measure respondent agreement with constituent item statements. We then aggregate these responses into corresponding indices using inverse covariance-weighting. Items are theoretically linked to corresponding, indexed concepts. Principal component analyses lend confidence to the theoretically motivated categorization scheme we employ by confirming items load on a common dimension.

Our survey did not include an experimental manipulation. Rather, we leverage a quasi-experiment posed by Hurricane Ian, which made landfall in Florida on September 28, 2022. Because the storm’s exact track and severity were determined by meteorological conditions, Ian constitutes a plausibly exogenous shock to attitudes. Balance and equivalence tests (Figures A-1 and A-2) bolster this claim, revealing few demographic differences between hurricane-exposed and unaffected respondents. These tests offer encouraging evidence against selective attrition, for instance, as a result of differential, hurricane-induced out-migration.Footnote ²¹

We define hurricane exposure at the county level using microdata on Ian’s eyepath, windswath, and storm surge (Figure 2), which we combine into an index.Footnote ²² We validate our exposure index in Figure 3, which shows our measure is highly correlated with self-reported hurricane exposure but not exposure to other climate disasters. A one-standard-deviation (sd) increase in exposure to Hurricane Ian increased self-reports of hurricane experience by 16–19 percentage points (pp). We also confirm that our hurricane exposure measure is associated with personal familiarity with hurricane-induced displacement (Table SI-4). After Hurricane Ian, respondents in affected counties became 3.2 pp more likely to report knowing someone who had moved because of a hurricane.

Figure 2. Mapping Hurricane Ian

Note: In panel (a), bins represent percentiles of the hurricane exposure index for values greater than the minimum of the index. The dashed red line marks the eyepath of Hurricane Ian.

Figure 3. Validating the Hurricane Exposure Measure

Note: Bars are 90% and 95% confidence intervals. Exposure is a continuous, z-standardized index combining information on Hurricane Ian’s eyepath, windswath, and storm surge. Estimations include covariates from Table 2. The dashed red line marks 0. Full tabular results are in Tables SI-2 and SI-3.

Our difference-in-differences approach compares respondent attitudes in exposed and unexposed counties before and after landfall. The identifying assumption is that in the absence of Hurricane Ian, exposed and unaffected counties would experience common trends in outcomes. In Figures SI-1, SI-2, and A-3, we provide graphical evidence of parallel pre-trends in event studies. That outcomes are consistently parallel in the pretreatment period builds confidence in the design. Formally, we estimate a least-squares equation:

$$ {Y}_{i,c,t}={\displaystyle \begin{array}{l}{\alpha}_c+{\beta}_t+\delta \left({\mathrm{Hurricane}\ \mathrm{Exposure}}_c\right)\times \left({\mathrm{Post}}_t\right)\\ {}+\hskip2px \gamma \left({X}_i\right)+\epsilon, \end{array}} $$

where i indexes respondents, c indexes counties, and t indexes the survey date. $ {Y}_{i,c,t} $ are climate attitudes, $ {\alpha}_c $ are county fixed effects, $ {\beta}_t $ are date fixed effects, and $ {X}_i $ is a vector of individual-level covariates. We interact $ {\mathrm{Hurricane}\ \mathrm{Exposure}}_c $ , a time-invariant measure of county-level storm severity, with $ {\mathrm{Post}}_t $ , an indicator for dates on or after landfall. Constitutive terms of the interaction do not appear separately because they are fully absorbed by county and time fixed effects. The coefficient $ \delta $ captures the extent to which Hurricane Ian induced a differential change in attitudes in exposed counties relative to counties unaffected by Hurricane Ian. This estimate represents the causal effect of Hurricane Ian on climate attitudes. $ \epsilon $ are heteroskedasticity-robust, county-clustered standard errors. Estimates are scaled using sampling weights.Footnote ²³

ROBUSTNESS

Returning to our main attitudinal results, we probe robustness in a variety of additional specifications, all of which corroborate the large, positive effect of Hurricane Ian on climate opinion. In Tables SI-5–SI-7 and A-4, we explore a number of alternative measures of hurricane exposure, including components of our index (e.g., windswath and storm surge) and a binary decomposition of the index, which takes a value of 1 for counties above the median and 0 otherwise. Whereas our main index captures the county-level intensity of Hurricane Ian, this binary measure averages over substantive, scale effects of hurricane severity.

Second, in Tables A-5 and A-6, we verify that similar results emerge using alternative outcome indices constructed by principal component analysis or by averaging constituent items. Third, we confirm that the estimates hold while matching hurricane-exposed and unexposed respondents on demographic covariates (Table A-7). Fourth, while the regressions include pertinent controls, we confirm that the estimated effects are robust to incorporating a broader array of covariates, like race and religiosity (Table A-8). Fifth, we find that the main effects are robust to alternative error clustering (Tables A-9 and A-10), sampling weights (Figure A-6), and estimators (Table A-11).

For omitted time-varying variables to bias our estimates, they must vary daily across counties. Three relevant confounders stand out: local politics, migration, and hurricane-induced displacement. We lack daily information on these covariates, so instead we draw on pretreatment measures. In Table A-12, we incorporate these pre-hurricane, county-level controls flexibly by interacting them with date fixed effects. To capture local politics, we take the county-level Republican voteshare from the 2020 Presidential election. To capture migration trends, we take 2021 county-level net migration rate. To capture hurricane-related displacement, we study data from Waze, a traffic-mapping application. Before Ian, Waze partnered with the Florida government to track evacuation-related road hazards. We use these data to estimate the population-normalized intensity of hurricane-induced traffic before landfall. Results are robust to accounting for these potential confounders.

Of course, while we account for many theoretically relevant factors, it remains possible that omitted variables could bias our results. We conduct several sensitivity approaches to assess the degree of confounding from unobservables that would be required to alter the substantive interpretation of our findings. Using a test proposed by Cinelli and Hazlett (Reference Cinelli and Hazlett2020), we benchmark confounding from unobservables against observed covariates. Figure A-7 reveals that even a confounder three times stronger than Democratic partisanship—the most important predictor of Americans’ climate attitudes—would be insufficient to alter the results. Table A-13 and Figure A-8 yield similar evidence using related tests (Blackwell Reference Blackwell2014; Oster Reference Oster2019). Together, these sensitivity analyses suggest unobserved confounding would have to be implausibly strong to alter the main conclusions.

A number of supplemental tests also extend our core analyses. In Table A-14, we exploit Hurricane Ida, the strongest hurricane of 2021, as a placebo. Hurricane Ida made landfall in Louisiana a year before our survey and caused significant damage from coastal Texas to the Florida Panhandle. Counties exposed to Ida should be similar to counties exposed to Ian, but we should not observe an effect of Ida, conditioning on exposure to Ian. Reestimating the core specifications while studying Ida exposure confirms this. Additionally, in Figure A-10, we consider a placebo survey outcome—support for strengthening the U.S. military. We expect this outcome to be unaffected by hurricane exposure and estimate support by repeating the focal specifications from Table 2. Encouragingly, Ian had no effect on this unrelated placebo outcome. Finally, in Figure A-11, we test whether the main effects decay with distance from Hurricane Ian’s eyepath. Although hurricanes affect large areas, their destructive power is greatest along the eyepath. Work by Hazlett and Mildenberger (Reference Hazlett and Mildenberger2020) finds climate disasters often have highly localized effects. We find that large, precise effects of Ian decay by 100–500 miles of distance from the eyepath.

HETEROGENEOUS EFFECTS

Do the effects of Hurricane Ian vary across demographic subgroups? Identifying how key traits moderate the impact of hurricane exposure is central for understanding the mechanisms by which climate disasters shape attitudes and for understanding how political coalitions for pro-climate policymaking might be formed after disasters. We preregistered tests for heterogeneous effects across many theoretically relevant dimensions and focus on two particularly crucial traits—partisanship and income—in Table 4.Footnote ³⁰

Table 4. Heterogeneous Effects of Hurricane Exposure on Climate Attitudes

Note: Robust, county-clustered standard errors are in parentheses. Post is an indicator for all dates on or after September 28, 2022 when Hurricane Ian made landfall in the United States. Exposure is a continuous, z-standardized index combining information on Hurricane Ian’s eyepath, windswath, and storm surge. Demographic covariates are partisanship, education, gender, and age. Estimates show the effect of Hurricane Exposure× Post in subsamples defined by the respective trait denoted in the panel title. Full tabular results are in Tables A-15 and A-18. * $ p<0.10 $ , ** $ p<0.05 $ , *** $ p<0.01 $ .

In the top panel of Table 4, we study how respondent partisanship shapes responsiveness to Hurricane Ian. To do so, we repeat the core specifications from Table 2, while subsetting the sample to Democrats and Republicans, respectively. Recent work suggests that disasters only induce pro-climate behavior in Democratic areas (Hazlett and Mildenberger Reference Hazlett and Mildenberger2020). In contrast, we find little systematic evidence of heterogeneity by individual partisanship. Republicans’ and Democrats’ attitudes on climate migration and climate change are equally responsive to hurricane exposure. The only heterogeneous effect we document is greater responsiveness of Republicans’ beliefs in climate science: whereas exposure to Hurricane Ian had virtually no effect on Democrats’ beliefs in climate science, a one-standard-deviation increase in exposure increased Republicans’ beliefs by 0.22 sd. This likely owes to a ceiling effect among Democrats, whose baseline belief in climate science is much greater than that of Republicans. In general, we find evidence that the pro-climate impacts of Hurricane Ian cross-cut partisanship.Footnote ³¹ Pro-environment views may be mobilized on both sides of the aisle following large-scale climate disasters.

In the bottom panel of Table 4, we study how respondent income shapes responsiveness to Hurricane Ian. For these tests, we split the sample at the median of income and repeat the core specifications from Table 2 for low- and high-income subsamples. Considering income differences in the effect of climate disasters is critical for three reasons. First, the world’s poor are disproportionately climate-vulnerable, facing heightened food insecurity and physical risk from disasters (Hallegatte et al. Reference Hallegatte, Bangalore, Bonzanigo, Fay, Kane, Narloch and Rozenberg2015). Second, as a consequence of their disproportionate vulnerability, low-income people face the greatest climate-related migratory pressures (Rigaud et al. Reference Rigaud, de Sherbinin, Jones, Bergmann, Clement, Ober and Schewe2018) but are least able to afford displacement to climate-resilient communities. For instance, during hurricanes, low-income individuals are less able to afford evacuating and more likely to rely on public shelters for housing assistance (Riad, Norris, and Barry Ruback Reference Riad, Norris and Barry Ruback1999). Third, as a result of sociotropic, nativist concerns, poor climate migrants face the staunchest backlash from receiving communities (Marotzke, Semmann, and Milinski Reference Marotzke, Semmann and Milinski2020).

We find that compared with high-income respondents, low-income respondents’ climate attitudes are consistently more responsive to hurricane exposure. A one-standard-deviation increase in exposure to Hurricane Ian increased low-income respondents’ pro-climate opinions by 0.12–0.26 sd across outcomes. The difference in effect sizes between low- and high-income respondents is large and distinguishable for six of seven indices. Together, these findings are strongly suggestive of rational retrospection. Climate disasters are most impactful in shaping beliefs of individuals most vulnerable to climate change, least able to afford moving from severely impacted areas, and most reliant on public assistance in the event of displacement.

EFFECT PERSISTENCE

Are the effects of hurricane exposure durable? Previous work suggests climate disasters have short-lived consequences (Egan and Mullin Reference Egan and Mullin2012; Konisky, Hughes, and Kaylor Reference Konisky, Hughes and Kaylor2016). As discussed above, one prominent explanation for temporal decay in disasters’ effects is recency bias (Kahneman, Slovic, and Tversky Reference Kahneman, Slovic and Tversky1982). Climate disasters may mobilize a rapid spike in pro-climate opinion in the short run before new developments attenuate their catalyzing effects, causing attitudes to revert to baseline levels. Yet Hurricane Ian was a devastating storm with lingering infrastructural impacts. The intensity of storm damage in Ian-affected counties could render the main effects durable.

Our original survey ran for 4 weeks after Hurricane Ian made landfall, allowing us to probe short-run effect persistence. Event study estimates (Figures SI-1 and SI-2) reveal that the main effects of interest all persisted for at least the first month after landfall. Our political behavioral results (Table 3) also comport with this finding. Florida voters in counties more severely impacted by Hurricane Ian were more supportive of pro-climate proposals 5 weeks after landfall.

To further assess effect persistence, we fielded an exploratory follow-up survey on a new sample of respondents (Figure A-12) in March 2023, roughly 6 months after Hurricane Ian. Results from this follow-up are depicted in Figure 5, alongside baseline estimates from Table 2. Corresponding with previous studies, we find that the effects of hurricane exposure attenuate in our follow-up survey.Footnote ³² Whereas the main effects were large and precise for at least 1 month after Hurricane Ian, all effects of hurricane exposure become null by 6 months post-storm. Our design is unable to reveal precisely when the effects of Hurricane Ian attenuated between our original and follow-up surveys, but the relatively short-lived effects we document accord with existing findings on climate attitudes (e.g., Egan and Mullin Reference Egan and Mullin2012) and recency bias (e.g., Nordhaus Reference Nordhaus1975). Probing how durable effects of disaster exposure are, and disentangling causal mechanisms, remains an important avenue for future research.

Figure 5. Effects of Hurricane Exposure in a 6-Month Follow-Up

Note: Bars are 90% and 95% confidence intervals. Exposure is a continuous, z-standardized index combining information on Hurricane Ian’s eyepath, windswath, and storm surge. Black markers denote estimates from the main sample (Table 2). Gray markers are coefficients from the follow-up sample fielded in March 2023 and represent the correlation between hurricane exposure and attitudes. Estimations include covariates from Table 2. The dashed red line marks 0. Full tabular results are in Table 2 (black estimates) and Table A-20 (gray estimates).

We also explore heterogeneity in effect persistence.Footnote ³³ Two especially relevant dimensions that could impact the durability of Ian’s effects are the occurrence of additional disasters and provision of post-disaster relief. To assess the first possibility, we study how exposure to a subsequent, late-season hurricane—Nicole—moderated Ian’s effects. Hurricane Nicole made landfall in Florida on November 10, 2022, 2 weeks after our initial survey ended. Although Nicole was a much weaker storm than Ian, it impacted similar areas, affecting portions of Florida and North Carolina (Figure SI-3). Multiple disaster exposure could magnify the effects of Hurricane Ian by reinforcing the pressing need for climate mitigation or blunt effects of Ian by distracting public attention (Arndt, Jensen, and Wenzelburger Reference Arndt, Jensen and Wenzelburger2021) and muting climate risk perceptions among doubly victimized populations (Leppold et al. Reference Leppold, Gibbs, Block, Reifels and Quinn2022). In Table SI-10, we find suggestive evidence of the latter. The persistence of Hurricane Ian’s pro-climate effects is greater among respondents exposed to Hurricane Ian but not Hurricane Nicole than among respondents exposed to both storms.

To test how post-storm aid shaped the persistence of Hurricane Ian’s effects, we assemble data on individual and public assistance distributed by the Federal Emergency Management Agency (FEMA) in areas impacted by Hurricane Ian. At the time of our follow-up survey, FEMA had already disbursed more than $1.5 billion in relief to Ian’s victims. We lack information on whether respondents themselves had received assistance but define an indicator for counties that had received federal aid by the time of our follow-up. Disaster relief may foster effect persistence by generating voter gratitude (Bechtel and Hainmueller Reference Bechtel and Hainmueller2011), or it may undermine effect persistence by inducing beneficiaries to focus myopically on recovery over mitigation (Healy and Malhotra Reference Healy and Malhotra2009). Disaster relief that enables victims to remain in their original communities, for instance, by funding home repairs, may also undercut specific support for policies to benefit climate migrants. Climate-induced migratory pressures are greater for individuals exposed to hurricanes and who have not received federal relief that could make staying feasible.Footnote ³⁴ Consistent with this logic, Table SI-11 reveals that the positive effect of Hurricane Ian on climate migration attitudes was longer-lasting for respondents in counties that had yet to receive federal relief. This finding underscores a tradeoff for pro-environment policymakers responding to disasters—providing relief aid improves victims’ welfare but potentially undermines the durability of disasters’ pro-climate effects.

We leverage rich demographic data from our follow-up survey—as in Table 4—to further consider whether effects are more durable among various population subgroups (Tables SI12–SI-16). These tests reveal scant evidence of heterogeneity in the persistence of Hurricane Ian’s effects across demographic traits. In sum, we find relatively short-lived effects of Hurricane Ian on victims’ climate attitudes. This suggests that while climate disasters open brief windows for policy action on climate migration and climate change, political opportunities are fleeting. Public officials interested in climate mitigation may be able to advance the pro-climate agenda by seizing on public favorability in the aftermath of disasters; however, policymaking timelines are often slower than disaster-induced surges of mass support. Discordance between electoral timetables and windows-of-favorability around disasters constrain prospects for major climate policy advances as a result of climate disasters. Still, pro-environment policymakers and candidates may be able to leverage favorable public opinion. As our behavioral results (Table 3) suggest, climate disasters can mobilize pro-climate voters when they occur in temporal proximity to elections. Policymakers should also take repeat disasters and post-disaster relief into consideration and target population subgroups for whom disasters have longer-lasting effects. Communities with single (rather than multiple) disaster exposure, and that do not receive post-disaster relief, represent one potential pro-climate coalition.