Disasters caused by extreme-weather events exacerbate social inequities. Already limited in resources and the ability to respond and recover from disasters, socially vulnerable people affected by disasters tend to suffer greater damages (Peacock et al. Reference Peacock, Van Zandt, Zhang and Highfield2014, 359–62) and poorer health outcomes (Smith et al. Reference Smith, Eshita Erica Anjum, Deanes and Acey2022, 83–84). This is due in part because they are pushed into less-desirable—and more hazard-exposed—areas (Bullard and Wright Reference Bullard and Wright2012, 50). Yet, disaster assistance does not always flow to the most vulnerable because it often is highly politicized and racialized (e.g., Garrett and Sobel Reference Garrett and Sobel2003, 507–508; Rivera-Burgos Reference Rivera-Burgos2023, 790). As climate change escalates the intensity and severity of natural hazards, greater cost of disasters (National Oceanic and Atmospheric Administration, National Centers for Environmental Information 2023), along with rapidly changing demographics (Bonilla-Silva Reference Bonilla-Silva, Back and Solomos2020, 652–56), will likely make the allocation of federal disaster relief increasingly contested. Previous research found that as the size and cost of government assistance programs increase—and recipients of such assistance increasingly are non-white and/or non-citizens—public support declines, especially among white and Republican voters (Filindra Reference Filindra2013, 29–32; Garand, Xu, and Davis Reference Garand, Xu and Davis2017, 155–56). To better understand this, we asked: “How do social identities affect perceptions of disaster-assistance deservingness?”
As climate change escalates the intensity and severity of natural hazards, greater cost of disasters, along with rapidly changing demographics, likely will make the allocation of federal disaster relief increasingly contested.
Building on welfare-policy scholarship, we present a theoretical framework for understanding who the public deems deserving of disaster assistance and apply it to two major hurricanes in 2017: Harvey and Maria. Although the devastation and loss of life caused by Hurricane Maria in Puerto Rico was worse than what was caused by Hurricane Harvey in Texas, the federal government responded faster and poured more resources into the Continental South than the commonwealth island (Willison et al. Reference Willison, Singer, Creary and Greer2019, 2). Using an original survey with an embedded experiment that was conducted in 2018, we explored how the social identities of race/ethnicity and partisanship are associated with attitudes about disaster-assistance deservingness.
Building on welfare-policy scholarship, we present a theoretical framework for understanding who the public deems deserving of disaster assistance and apply it to two major hurricanes in 2017: Harvey and Maria.
THEORETICAL FRAMEWORK FOR DISASTER-ASSISTANCE DESERVINGNESS
Disaster assistance has many similarities to welfare policies. It is driven by need; typically takes the form of a direct, nontaxable grant; and carries racial and redistributive effects (Emrich Aksha, and Zhou Reference Emrich, Aksha and Zhou2022, 15; Haselswerdt Reference Haselswerdt2022, 273–74). From differences in how whites and Blacks rate the assistance deservingness of Hurricane Katrina evacuees (Huddy and Feldman Reference Huddy and Feldman2006, 105) to the “spillover” of racial attitudes into federal-agency evaluations after Hurricane Sandy (Sheagley, Chen, and Farhart Reference Sheagley, Chen and Farhart2017, 108–12), disaster assistance also demonstrates similar trends in the partisan- and race-based opposition aimed at redistributive welfare policies. In 2017, President Trump pitted Puerto Ricans against Texans by tweeting that “yet more money” to this “place” (i.e., Puerto Rico) is “taking dollars away” from others.
Building on the US welfare-policy scholarship, we posit that salient social identities of race/ethnicity, nationality, and partisanship shape beliefs about disaster-assistance deservingness. Welfare-policy scholars found that public support and attitudes of deservingness are contingent on various social identities but are most closely tied to race and ethnicity (Filindra Reference Filindra2013, 29–32; Hancock Reference Hancock2004, 23–35; Marchevsky and Theoharis Reference Marchevsky and Theoharis2006, 1–12). These identities are fundamental to attitudes about redistributive welfare policies because racial and ethnic minorities in the United States are overrepresented among the poorest Americans; therefore, redistribution measures disproportionately benefit these groups. Researchers believe in-group/out-group biases result from the tendency for people to evaluate groups by those who are not members of their own group (Mason Reference Mason2018, 7–15). As such, racial/ethnic out-groups are viewed more negatively than one’s own racial/ethnic group in order to maintain a positive sense of social identity (Tajfel and Turner Reference Tajfel, Turner, Worchel and Austin1986, 7–16; Worchel et al. Reference Worchel, Morales, Paez and Deschamps1998, 53–60) and/or to express prejudicial devaluation of the out-group, aligned with one’s own minority bias (Vala, Pereira, and Costa-Lopes Reference Vala, Pereira and Costa‐Lopes2009, 22–23). Stated another way, group affiliation conditions messaging—that is, group members support the message of the in-group because it often reaffirms their own views and beliefs, and they reject out-group messaging, regardless of facts (Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012). This process is represented by the idea of “deservingness.”
Once primarily targeted at African Americans, the frame of undeservingness was extended to Latinos and immigrants during debates about welfare reform in the 1990s (Filindra Reference Filindra2013, 30–31). Research found polarization among the US public in support of state services for immigrants as perceptions about entitled citizens versus non-entitled non-citizens prevail (Filindra Reference Filindra2013, 30–31; Garand, Xu, and Davis Reference Garand, Xu and Davis2017, 151). The causal mechanisms that explain perceived undeservingness of immigrants are similar to—and increasingly overlap with—those mechanisms driven by racial animosity because the majority of current immigrants to the United States are from Latin America and Asia. There is a close association in political rhetoric between immigrants and racial “others” (Pérez Reference Pérez2015, 550–51).
Furthermore, how the public regards the type of intervention by the government (i.e., direct assistance or indirect market incentives) also matters for beliefs about deservingness (Gainous, Craig, and Martinez Reference Gainous, Craig and Martinez2008, 980; Haselswerdt and Bartels Reference Haselswerdt and Bartels2015, 608–10). Partisanship is the driver of attitudes about government intervention because more conservative (i.e., Republican) voters tend to have stronger animosity toward distributive politics (Goenka and Thomas Reference Goenka and Thomas2022, 313; Parker and Barreto Reference Parker and Barreto2015, 20–25). Also, because partisanship has become a salient social identity whereby people associate with their in-groups and show animus toward out-groups (Mason Reference Mason2018, 14–16), partisanship influences attitudes about who deserves assistance, not only what type of assistance (Hussey and Pearson-Merkowitz Reference Hussey and Pearson-Merkowitz2013, 574). Based on these theoretical insights, we offer the following two hypotheses:
Perceived Citizenship Hypothesis (H1): The perception that the person impacted by the disaster is a non-citizen will decrease support for more-generous and direct federal assistance.
Partisanship Hypothesis (H2): Republicans will be less likely than Democrats to support more-generous and direct federal disaster assistance.
DATA AND METHODS
The data for this study were drawn from a probabilistic, national survey (i.e., the University of Maryland Critical Issues Poll) conducted by Nielson Scarborough from October 24 to November 16, 2018. The sample included 1,300 total respondents, with an oversample of 600 Hispanics and 700 non-Hispanics. The margin of error for the survey was +/-2.72%.
To test our hypotheses, we developed an original experiment, embedded in the survey (Ross et al. Reference Ross, Rouse, Alcañiz and Marchevsky2024). Survey respondents were randomly assigned to a vignette about a Hurricane Maria– or a Hurricane Harvey–affected individual, varied by race/ethnicity and gender. Central to this experiment was the inclusion of Puerto Ricans, who often are perceived as less American than citizens on the US mainland. Therefore, they may be subject to perceptions of undeservingness, despite suffering widespread and catastrophic losses and damages as a result of Hurricane Maria (Rivera-Burgos Reference Rivera-Burgos2023, 2; Valle Reference Valle2018, 29). With an estimated 2,975 fatalities (Pasch, Penny, and Berg Reference Pasch, Penny and Berg2023) and more than $94 billion in storm damages (Kishore et al. Reference Kishore, Marqués, Mahmud and Kiang2018), Hurricane Maria is the fourth-costliest storm on national record (National Oceanic and Atmospheric Administration, National Centers for Environmental Information 2023). The damage included devastation of Puerto Rico’s energy grid; cascading failures of transportation, communications, water and wastewater infrastructure; and health and social services that overburdened an already socially vulnerable population (Chandra et al. Reference Chandra, Marsh, Madrigano, Simmons, Abir, Chan, Ryan, Nanda, Ziegler and Nelson2021). In comparison, Hurricane Harvey was responsible for approximately 100 deaths and an estimated $51 billion in damages in Texas (Blake and Zelinsky Reference Blake and Zelinsky2018).
Due to national attention on the severity of Hurricane Harvey and Hurricane Maria, a control group was not included in the experiment because it would be implausible for participants to disassociate disaster assistance with one of these events. The experiment randomly assigned survey respondents a name of a fictional disaster victim—designed to imply race/ethnicity as well as gender—who was forced to relocate due to the destruction of their property as a result of Hurricane Harvey or Maria. The following vignette was presented to respondents:
[Name] lost [his/her] house to flooding in Hurricane [Harvey/Maria] and had to relocate from the [city of Houston/island of Puerto Rico] to [a nearby city/the US mainland]. Should the federal government, through the Federal Emergency Management Agency (FEMA) provide funding to buy a house of similar value, provide loans to buy a house of similar value, subsidize a rental for an indefinite time, subsidize a rental for 6 to 12 months, or not offer any assistance?
Survey participants were distributed evenly across 16 treatment conditions, as shown in figure 1.
Figure 1 Experimental Group Assignments
The dependent variable of interest—that is, preference for disaster-assistance type—was measured by responses to the experimental vignette, including no assistance, subsidize a rental for six to 12 months, subsidize a rental for an indefinite time, provide a loan to buy a house of similar value, and funding to buy a house of similar value. These responses varied by generosity, with no assistance on one extreme and funding to buy a house on the other. They also varied by directness of federal assistance with funding for house and rental subsidies, typically providing direct benefits to a disaster affected individual, whereas a home loan provides indirect market-based incentives. These disaster-assistance types correspond to housing assistance provided by FEMA for primary residences damaged by a federally declared disaster, which includes initial and continued rental assistance (Federal Emergency Management Agency 2019) and home replacement grants up to $25,000 for homeowners who were uninsured or underinsured (Federal Emergency Management Agency n.d.).
To test the Perceived Citizenship Hypothesis (H1), we relied on the location cues of the experimental vignette assignment. Each respondent was assigned either Hurricane Harvey, which affected people in the city of Houston, or Hurricane Maria, which affected people on the island of Puerto Rico. Based on evidence from public polling, we assumed that these location cues represented perceived citizenship. In 2016, a poll conducted by Economist/YouGov of 2,000 American adults (+/-3% margin of error) found that only 43% of respondents knew that Puerto Ricans are American citizens. We expected Hurricane Maria to be associated negatively with more-generous and direct disaster assistance because it implies benefits to those who might be perceived as non-citizens. The Partisanship Hypothesis (H2) was tested by including the respondent’s political party affiliation, with the expectation that Republicans are negatively associated with more-generous and direct disaster assistance.
Other factors may influence attitudes about disaster assistance. As discussed previously, scholars found that public support for welfare assistance is closely tied to race and ethnicity (Filindra Reference Filindra2013, 29–32; Hancock Reference Hancock2004, 23–35; Marchevsky and Theoharis Reference Marchevsky and Theoharis2006, 1–12). Accordingly, we included measures of a respondent’s race/ethnicity and parental country of origin. We also accounted for risk perception—measured as disaster experience and perceived community risk to a disaster—because years of research found that these shape disaster attitudes and behaviors (Wachinger et al. Reference Wachinger, Renn, Begg and Kuhlicke2013, 1051). Additionally, attitudes about government (versus individual) responsibility may influence support for disaster assistance (Ross and Atoba Reference Ross and Atoba2022, 5). We used two measures to capture this attitude: (1) beliefs about government (versus individual) responsibility for disaster recovery, and (2) the likelihood of filing a claim with FEMA in the event of a disaster. We also included sociodemographic characteristics of the respondents, including gender, age, education, and income, because they also may affect disaster-assistance perceptions. (See online appendix A for coding of variables and online appendix B for descriptive statistics.)
We controlled for the other factors that may shape attitudes about disaster deservingness through regression analysis. Although the dependent-variable categories appear ordered in generosity of assistance, statistical tests indicate that ordinal logistic regression is not appropriate because the parallel regression assumption is violated (Long and Freese Reference Long and Freese2006, 199). Therefore, we estimated a multinomial logit regression, designating the dependent-variable outcome of “no assistance” as the base category. This allowed a comparison of the likelihood of preference for each disaster-assistance type to the preference of no assistance. A survey weight was applied to adjust the sample to national population characteristics. Due to missing data, a total of 1,111 observations were included in the model.
RESULTS
Survey respondents polled for this study were supportive of providing government disaster assistance: 40% supported a loan for a house, 30% preferred six to 12 months rental assistance, 14% supported the granting of funds to buy a house of similar value, and 7% preferred indefinite rental assistance. Only 9% stated that FEMA should provide no assistance to people affected by Hurricane Harvey and Hurricane Maria. The results of the multinomial regression estimation, reported in table 1, indicate that perceived citizenship and partisanship were coupled with attitudes about disaster assistance. However, the statistically significant association of these variables differed across disaster-assistance types. Given that we designated the dependent-variable outcome of “no assistance” as the baseline, coefficient signs were interpreted relative to this response. Because logistic coefficients are difficult to interpret directly, marginal effects also are reported in the table. They should be interpreted as the change in the likelihood of the dependent-variable outcome associated with the variable compared to the base category of that variable.
Table 1 Factors Associated with Federal Government Disaster Assistance Preferences
Notes: Multinominal logit regression estimated with base category of “no assistance”; coefficients reported in “Coeff” column with standard errors shown below in parentheses; marginal effects (dydx) reported in “ME” column with comparisons made to base category of the variable (see online appendix A for variable coding); statistical significance denoted as **p<0.01, *p<0.05.
The results indicate that assignment to a Hurricane Maria vignette was statistically significant for three disaster-assistance types: rent indefinitely, a loan for a house, and funds to buy a house. Marginal effects indicate that a respondent assigned to Hurricane Maria (rather than Hurricane Harvey) had, on average, a 2% lower likelihood of supporting rent indefinitely and a 5% lower likelihood of supporting a loan or funds to buy a house. To explore whether the assignment of Hurricane Maria was cuing Latino animosity, we conducted additional analyses (see online appendix C) that replaced the disaster-event variable with one that indicated assignment of a Latino disaster-affected individual (i.e., for Hurricane Maria, for Hurricane Harvey, or for both events). The results were statistically insignificant, suggesting that Hurricane Maria was not prompting Latino animosity but rather was cueing perceptions about the event itself, which we assert are related to attitudes about the place and the people who live there. This lends support for the Perceived Citizenship Hypothesis that those affected by Hurricane Maria (i.e., Puerto Ricans) were deemed less deserving of assistance, presumably because they are perceived as non- or less American.
We also found that partisanship was associated with disaster-assistance attitudes. The variable “Republican” was negative and statistically significant for the outcomes of six to 12 months rental assistance and a loan for a house. Marginal effects indicated that a respondent who identified as a Republican, on average, was 9% less likely to support six to 12 months of rental assistance than a respondent who identified as a Democrat or an Independent. The marginal-effect difference for a home loan was not significant (i.e., less than 1%), which indicates that partisanship primarily was associated with differences in attitudes about short-term rentals.
Marginal effects estimated for statistically significant control variables also demonstrate that a respondent with both parents born in the United States had, on average, a 6% and a 7% lower likelihood of supporting short-term rental assistance and a loan for a house, respectively, than a respondent with one or both parents born outside of the country. These respondents also were slightly more likely (2%) to support funds for a house. The results indicate that a Latino respondent had, on average, an 11% and a 17% lower likelihood of preferring short-term rental assistance and a loan for a house, respectively, than a non-Latino. Black, Asian, and respondents of other races and ethnicities had a 10% and a 27% higher likelihood of supporting indefinite rental assistance and funds to buy a house, respectively. Together, these results suggest that minorities and individuals with parents born outside of the United States favored more-generous forms of disaster assistance.
Although the two variables that measured risk perceptions (i.e., disaster experience and community disaster risk) were not statistically significant, the perception that the government is responsible for disaster recovery and stating that one is likely to file a claim with FEMA in the event of a disaster are significantly associated with disaster-assistance attitudes. Marginal effects demonstrate that respondents who believed the government is responsible for disaster response and recovery had, on average, a 15% lower likelihood of preference for short-term rental assistance. Those respondents also had a 3%, a 7%, and a 12% higher likelihood of supporting indefinite rental assistance, a loan for a house, and funds for a house, respectively, compared to respondents who believed that it is an individual’s responsibility to respond and recover from a disaster event. Similarly, respondents had, on average, a 2% and a 4% higher likelihood of supporting a loan and funds for a house, respectively. Their self-reported likelihood of using FEMA assistance in the event of a disaster increased from “not likely” to “not very likely,” “somewhat likely,” and “very likely.” These findings suggest that those respondents who believed the government is responsible for disasters—and could see themselves relying on such aid—preferred more-generous disaster assistance.
Female gender, age, and education also demonstrated a statistically significant association with disaster attitudes. Marginal effects suggest that a female respondent (compared to a male) had, on average, a 15% lower likelihood of supporting a loan for a house. Regarding age, a respondent had, on average, less than 1% and about a 2% higher likelihood of preferring short-term rental assistance and a home loan, respectively, for each increase in the age category from the base of 18 to 20 years. Age also was associated with an almost 1% lower likelihood of support for funds to buy a house; this implies that older adults prefer less-generous and market-based forms of assistance.
The discussion now considers how citizenship perceptions, partisan leanings, and respondent race/ethnicity intersect to influence attitudes about disaster-assistance deservingness. Specifically, we estimated the marginal effects of the common covariate profile for each race and ethnic group on disaster-assistance support. The values were assigned based on the most frequent category or the mean of each variable (see online appendix D). The common profile for white respondents was a male who is older, more educated, and wealthier than the survey sample (on average); is a Republican with both parents born in the United States; has no disaster experience and low concern for community disaster risk; believes the government is responsible for disaster response/recovery; and is likely to file a claim with FEMA if affected by a disaster. The common profile for Latino respondents was a female who is younger, less wealthy, and less educated than the survey sample; is a Democrat or Independent with one or both parents born outside of the United States; has no disaster experience but high concern for community disaster risk; believes the government is responsible for disaster response/recovery; and is likely to file a claim with FEMA if affected by a disaster. The common profile for a respondent of other race/ethnicity also was a female who is younger and slightly less wealthy but more educated than the survey sample; is a Democrat or Independent with both parents born in the United States; has no disaster experience but high concern for community disaster risk; believes the government is responsible for disaster response/recovery; and is likely to file a claim with FEMA if affected by a disaster. We calculated the marginal effects of these common profiles for each vignette assignment (i.e., Hurricane Maria or Hurricane Harvey).
The marginal effects of the common profiles, reported in table 2, demonstrate that the intersection of race/ethnicity, partisanship, and disaster location perceptions was associated with substantial differences about disaster assistance and deservingness. Profiles 1 and 2 underscore that white Republicans were most likely (60% to 65%) to support market-based disaster assistance in the form of a home loan. They also were slightly more likely (approximately 5%) to favor a home loan for Hurricane Harvey–affected people than those affected by Hurricane Maria. Profiles 3, 4, 5, and 6 highlight that Latino and other minorities who affiliated with the Democratic Party or were an Independent were most likely to prefer the most-generous form of assistance—funds for the purchase of house—ranging from approximately 37% among Latinos to 58% among other minorities. The likelihood of support for a home loan follows close behind the race/ethnic minority profiles, with the highest likelihood (30%) of preferring this form of assistance among Latinos who were assigned a Hurricane Maria vignette.
Table 2 Likelihood of Support for Disaster Assistance across Common Covariate Profiles
Notes: Marginal effects estimated using the multinominal regression results in table 2; boldfaced percentages indicate the highest likelihood for that profile; see online appendix D for the values set for each covariate in the model.
Together, these marginal-effects profiles underscore that the location of the disaster event—that is, the island of Puerto Rico (Hurricane Maria) or the city of Houston, Texas (Hurricane Harvey)—influenced deservingness of disaster assistance because all profiles tended to support less-generous forms of assistance for Hurricane Maria–affected people and more-generous aid for Hurricane Harvey–affected people. These findings begin to unravel the complexities of intersecting social identities in understanding attitudes about disaster-assistance deservingness that should be explored further in future studies.
DISCUSSION AND CONCLUSION
This study examines how questions about deservingness—conditioned by social identities—affected support for federal disaster assistance. Using a 2018 survey experiment that leveraged two devastating hurricanes—Hurricane Maria that hit Puerto Rico and Hurricane Harvey that hit the southern part of the Continental United States in 2017—we examine how the social identities of race/ethnicity and partisanship affect attitudes about disaster deservingness. We find that federal disaster assistance has broad support. This implies that disaster assistance is regarded by the public as a form of social insurance that protects against risk rather than as government assistance programs that are perceived as welfare (Meyer Reference Meyer2013, 329). However, the public’s support for disaster assistance is not uniform or unconditional but instead varies depending on perceptions about the disaster victim and the type of assistance. Survey respondents assigned to a vignette featuring a disaster-affected person from Puerto Rico were less willing to support the most-generous and direct forms of disaster assistance. Furthermore, support for disaster assistance varied across respondents’ party affiliation and personal race/ethnicity; the highest likelihood of preference for a home loan was among white Republicans, whereas Latinos of all political parties were most likely to express support for the most-generous form of assistance: that is, funds for a house—albeit less for those assigned to a Hurricane Maria vignette.
Together, our findings suggest that when the US public perceives people who are affected by disasters to be non- or lesser citizens, it views them as less deserving and supports less-generous forms of government assistance to these groups. As climate change intensifies and the frequency of billion-dollar disasters continues to grow, American attitudes about deservingness likely will have greater consequences, particularly if Congress restructures disaster-relief spending (Painter Reference Painter2020, 38). In the face of an increasingly polarized and racialized social and political environment, these dynamics could have severe negative consequences for racial and ethnic minorities as well as immigrants, thereby hampering how disaster assistance reaches the most vulnerable and needy (Donner and Rodríguez Reference Donner and Rodríguez2008, 1089–92; Hansen et al. Reference Hansen, Bi, Saniotis and Nitschke2013, 1–4).
[O]ur findings suggest that when the US public perceives people who are affected by disasters to be non- or lesser citizens, it views them as less deserving and supports less-generous forms of government assistance to these groups.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit http://doi.org/10.1017/S1049096524000362.
ACKNOWLEDGMENTS
We thank Shibley Telhami for feedback on the development of the survey experiment. We also thank the anonymous reviewers for providing comments that significantly improved this article. Any errors or omissions remain our own.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the PS: Political Science & Politics Harvard Dataverse at https://doi.org/10.7910/DVN/XIXD0J.
CONFLICTS OF INTEREST
The authors declare that there are no ethical issues or conflicts of interest in this research.
Open access
