(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Disparities in self-reported extreme weather impacts by race, ethnicity, and income in the United States [1]

['Chad Zanocco', 'Civil', 'Environmental Engineering', 'Stanford University', 'Stanford', 'California', 'United States Of America', 'June Flora', 'Hilary Boudet', 'School Of Public Policy']

Date: 2022-06

Abstract Extreme weather events are expected to increase in frequency and severity due to climate change. However, we lack an understanding of how recent extreme weather events have impacted the U.S. population. We surveyed a representative sample of the U.S. public (n = 1071) in September 2021 about self-reported impacts they experienced from six types of extreme weather events within the past three years. We find that an overwhelming majority (86%) of the U.S. public reported being at least slightly impacted by an extreme weather event, and one-third (34%) reported being either very or extremely impacted by one or more types of extreme weather events. We clustered respondents into four impact groups, representing a composite of self-reported impacts from multiple types of extreme weather events. Respondents in the highest extreme weather impact group are more than 2.5 times as likely to identify as Black or Hispanic and 1.89 times more likely to live in a household with income levels below the Federal poverty level. We also observe reports of higher extreme weather impacts from respondents who are female, do not have a bachelor’s degree and live in a rural area. Our results indicate that extreme weather impacts are being felt by a broad cross-section of the U.S. public, with the highest impacts being disproportionately reported by populations that have previously been found to be more vulnerable to natural disasters and other extreme events.

Citation: Zanocco C, Flora J, Boudet H (2022) Disparities in self-reported extreme weather impacts by race, ethnicity, and income in the United States. PLOS Clim 1(6): e0000026. https://doi.org/10.1371/journal.pclm.0000026 Editor: Alessandra Giannini, École Normale Supérieure: Ecole Normale Superieure, FRANCE Received: November 18, 2021; Accepted: May 15, 2022; Published: June 16, 2022 Copyright: © 2022 Zanocco et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: Data applied in this study is available through the Open Science Framework and can be accessed at https://osf.io/7h6xv/. Funding: Support for this research was provided by the National Science Foundation’s Smart & Connected Communities Program (#1737565) (HB). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors declare that they have no competing interests.

1. Introduction The United Nations Secretary General described the recent IPCC Working Group report on climate change as a “code red” for humanity [1]. Moreover, the consequences of climate change are not equally distributed and are likely worsening global inequality [2–4], a trend that is expected to continue without comprehensive climate mitigation and adaptation policies. Such disparities can be observed through the differential impacts of extreme weather events, the frequency and severity of which—for certain events and under certain circumstances—can be linked to human caused climate change [5–7]. For example, at a global level, less developed countries are more likely to be impacted by increasingly frequent and intense heat waves due to climate change [8]. There are also notable heterogenous impacts among populations within countries. Such variation in population exposure and impacts is prominent in the United States [9], both for reasons related to geography but also due to existing social inequities, with certain populations (e.g., those living in poverty) being more vulnerable to the impacts of natural disasters [10]. The events themselves are not evenly distributed geographically. Certain extreme weather event types are more likely to occur within particular U.S. regions: hurricanes in the South and the East Coast, tornadoes in the Midwest, and wildfire in the West. Within and across these U.S. geographies, however, are differential climate-related impacts on populations, as well as a patchwork of climate adaptive policies applied at different levels of government [11]. An understanding of self-reported extreme events impacts at a population-level, therefore, is needed to characterize overall experiences of extreme weather across the U.S. public—and what members of society report being impacted the most. This is especially important given the constantly changing dynamics of extreme weather event impacts on U.S. populations and how these changing dynamics may intersect with impacts from other concurrent events, such as the COVID-19 pandemic [12]. Moreover, recent events may be impacting U.S. populations in ways that diverge from historical experiences with extreme weather—with 2021 alone including record-breaking wildfires [13], heat waves [14], winter storms [15], and hurricanes [16]—that left millions without power and water, sometimes for weeks at a time. 1.1 Experiences with extreme weather events Extant research that has examined experience, exposure, and harm from extreme weather events have often focused on individual events (e.g., [17, 18]), events of the same type or in the same area (e.g., [19, 20]), or a collection of selected events with a similar magnitude of impacts on communities [21, 22]. Such approaches have yielded insights into the characteristics of those who are impacted by specific extreme event types and their harm from these events (e.g., [23, 24]). However, studies with national-level coverage of the U.S. general public’s reported experiences with multiple extreme event types are less common (see [25] for an exception). Recent polling in Fall 2021 found that approximately half (46%) of Americans reported first-hand experiences with extreme weather within the past 12 months and that two-thirds (67%) believed that extreme weather events were happening more often than in the past [26]. However, this polling effort did not measure the severity of impacts from extreme weather events and did not delineate experiences by different event types. To address this gap, we examine how experience with extreme weather has recently impacted the U.S. public across a range of event types and geographies, formalized in the following research questions: RQ: What recent extreme weather events impacts are being self-reported by the U.S. public? How do these reported impacts vary by event type and geographic region? These research questions, however, do not provide insight into how specific populations within the U.S. may be differentially impacted by extreme events, especially among those groups who have previously been found to be more impacted by natural disasters, extreme weather, and other disruptive events. Therefore, we also examine self-reported extreme weather impacts by sociodemographic characteristics identified in previous literature as being associated with higher levels of impact. 1.2 Differential experiences with extreme event impacts Understanding the characteristics of those severely impacted by extreme weather is critical for planning, policy, and response efforts. Previous literature has established that certain populations can be more impacted by a variety of natural disasters and disruptive events, ranging from extreme weather events and environmental hazards to economic crises and pandemics [10, 27–29]. Using previous literature as a guide, we form a set of hypotheses with the overall objective of understanding disproportionate impacts from recent extreme weather reported by different populations in the United States. We focus on six interconnected sociodemographic characteristics that have been found in previous studies to relate to extreme event impacts: income (H1); race/ethnicity (H2 & H3); gender (H4); education (H5); age (H6); and rurality (H7). Low income populations and those living in poverty have been found to be disproportionately impacted by natural hazards and disasters, both across the world and in the U.S. [30, 31]. Explanations for this relationship between poverty and more negative impacts from extreme events are multifaceted, ranging from the location of lower income housing in more disaster exposed areas [32, 33], to disproportionate losses in assets and well-being due to events [34], to less capacity to immediately respond to events and fully recover post-event [35]. There is also evidence that climate change impacts can increase future poverty rates [36]. Given this relationship between disasters and poverty, we form the following hypothesis: H1. Respondents living in poverty will be more likely to report negative impacts from extreme weather events. The COVID-19 pandemic has underscored how a disruptive event can also have differential impacts on racial/ethnic minority groups, particularly in the U.S. [37–39]. Indeed, previous research has found that other types of disruptive events, such as natural disasters and extreme weather events, can have differential impacts by race and ethnicity [40, 41]. For example, research has found that phenomena such as extreme heat exposure via heat island effects disproportionately impacts neighborhoods with high percentages of Latinos and African Americans [42–44]. Additionally, in disasters such as Hurricane Katrina, more than half of the fatalities were Black [45]. And, during the 2021 winter storm in Texas, it was historically marginalized minority neighborhoods that were the first to experience power outages [46, 47]. We therefore propose the following hypotheses: H2. Respondents identifying as Black will be more likely to report negative impacts from extreme weather events. H3. Respondents identifying as Hispanic will be more likely to report negative impacts from extreme weather events. Research has also found that older adults are more likely to be impacted by disasters due to multiple factors, including a higher likelihood of chronic health conditions and sensory, cognitive, and mobility disabilities, as well as limited social and economic support systems [48, 49]. Evidence suggests that older adults have higher rates of mortality related to disasters [50], brought into sharp focus in the aftermath of California’s 2018 Camp Fire where three-quarters of victims were age 65 years or older [51]. We therefore propose the following hypothesis: H4. Older adults will be more likely to report negative impacts from extreme weather events. Gender has long been linked to differential disaster impacts, with women having less access to resources such as health care and emergency housing following extreme weather events and disasters [52]. Women have also been shown to be less likely to receive economic assistance following disasters, making recovery from extreme events more challenging [53]. Additionally, there are indications that domestic violence may increase in the wake of nature disasters [54], and, at a global level, there is a gendered gap in life expectancy, with life expectancy of women lowered more compared to men following natural disasters [55]. Given this relationship between gender and differential impacts from extreme weather events and disasters, we form the following hypothesis. H5. Females will be more likely to report negative impacts from extreme weather events. Having lower levels of educational attainment has also been found to be associated with increased impacts from extreme events as it can make preparing for disasters, accessing resources, and seeking out information and assistance during and after the event more challenging [56]. In turn, having higher levels of educational attainment, both at the individual-level and across society, is related to better preparedness and responses to disasters, as well as less negative impacts from events [57–59]. Given this connection between educational attainment and vulnerability, we pose the following hypothesis: H6. Respondents with lower educational attainment will be more likely to report negative impacts from extreme weather events. Residents living in rural communities may be more vulnerable to disasters and extreme events due to lack of local government capacity, as well as often less availability of government funding for prevention and recovery compared to urban settings [60, 61]. Moreover, an analysis of rural communities found them to be more vulnerable to climate change impacts—including extreme weather events—compared to urban areas, with high levels of variability by place and region [62]. And for rural communities dependent on agriculture, climate impacts can be multiscale, impacting both economic livelihoods and social processes [63]. Given this relationship between rural areas and increased potential for experiencing extreme weather impacts, we form the following hypothesis: H7. Respondents living in rural areas will be more likely to report negative impacts from extreme weather events. To answer our research questions and test our hypotheses, we surveyed a representative sample of the U.S. public and asked respondents to report their level of negative impacts from six extreme event types in the past three years. The remainder of this article is summarized as follows. First, we describe our survey data and analytical methods used to identify groups of respondents negatively impacted by extreme weather. Next, we provide a summary of self-reported extreme weather events impacts across the general population as well as provide descriptions of types of extreme events experienced across different regions of the United States. We then present findings on the relationship between our selected sociodemographic characteristics and negative impacts from extreme events. Finally, we interpret our findings and identify limitations, as well as discuss policy implications and proposed future research.

2. Data and methods 2.1 Data collection We use survey data collected through the AmeriSpeak bi-weekly Omnibus panel of the U.S. public, funded and operated by NORC at the University of Chicago. Survey participants were recruited from a probability-based panel that was randomly generated via address-based sampling and designed to be representative of the U.S. household population. To recruit survey participants for the AmeriSpeak panel, NORC contacts households by U.S. mail, telephone, and using field interviewers that conduct in-person recruitment. This results in panels with coverage of approximately 97 percent of the U.S. household population. Those populations who are more likely to be excluded from the sample have P.O. box only addresses, addresses that are not included in the USPS Delivery Sequence File, or live in newly built dwellings. For more detailed information about the AmeriSpeak panel and methodology, see [64] and for examples of recent research that applies survey data acquired from AmeriSpeak panels, see [65–67]. Our survey was administered online and by phone to 1071 participants who were 18 years or older from September 10–12, 2021 and included questions about extreme weather event impacts alongside other sociodemographic questions. Probability-based samples of this size are adequate for generating general population estimates and have even been shown to outperform much larger samples acquired using convenience-based sampling [68]. The overall margin of error for our survey sample was +/- 3.93 percentage points at the 95 percent confidence level. See Table J in S1 Text for a comparison between our survey sample composition and national U.S. population estimates. This study was reviewed by the Oregon State University Human Research Protection Program (IRB-2019-0137) and received an exempt determination. NORC, the survey vendor, obtained written informed consent from study participants and de-identified all data prior to transmission to the authors. 2.2 Survey measures Self-reported impacts from extreme weather events, our primary measure in this research, is formed from the multi-item survey question “How negatively impacted were you by you the following extreme events in the past three years?” with response categories 1 = “Not at all impacted”, 2 = “Slightly impacted”, 3 = “Moderately impacted”, 4 = “Very impacted” and 5 = “Extremely impacted” for six extreme event types or type combinations. These six extreme event item descriptions include “Extreme cold / winter storms” (mean = 2.059, std. dev. = 1.162), “Extreme heat” (mean = 2.475, std. dev. = 1.264), “Extreme rainfall / flooding” (mean = 1.815, std. dev. = 1.055), “Hurricanes” (mean = 1.551, std. dev. = 0.978), “Wildfires” (mean = 1.674, std. dev. = 1.101), and “Tornadoes” (mean = 1.449, std. dev. = 0.911). When combined into a single mean composite index, average negative impacts from event types are 1.841 (std. dev. = 0.759), a value closest to the “Slightly impacted” category. For a reproduction of the original survey question, see Box A in S1 Text. We acknowledge that conceptualizations of self-reported impacts are subjective and could include everything from the severity and frequency of negative experiences to different types of impacts including monetary and health impacts. In this research we sought to balance impact specificity and generalizability, seeking to develop an overall understanding of impacts using a parsimonious set of survey items that anyone in the U.S. can respond to. For examples of survey research that asked about more specific impacts, such as physical harm, from single named events, see [22, 24]. We also assess how these self-reported impacts relate to sociodemographic characteristics, specifically gender, race/ethnicity, household income, education, age, and rurality, described in detail in Table 1. Demographic characteristics are formed into the following variables: female (= 46.4% vs. male), 65 or older (= 22.1%), bachelor’s degree or higher (= 35.4%), Black (= 11.5%), Hispanic (= 16.4%), and Below Federal poverty level (= 19.7%). To construct the poverty measure, we first used household size to assign a Federal Poverty Level threshold to each respondent [69] and then compared this Federal Poverty Level threshold to the respondent’s reported household income. If the respondent’s reported household income was below this threshold, they were identified as being below the Federal poverty level. Additionally, we also apply a measure of respondent residence in a rural area using the designation of non-metropolitan (= 18%). Lastly, we apply information about respondents’ locations within Census Regions (n = 4) and Census Divisions (n = 9) (see Table 1 for more detail). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Table 1. Summary of measures, descriptions, descriptive statistics, and variables applied in this research. https://doi.org/10.1371/journal.pclm.0000026.t001 2.3 Analytical approach We apply a multi-stage analytical approach in this research. After generating descriptive statistics for extreme weather event impacts across types and geographies, in our first stage of analysis we identify groups of respondents with similar levels of self-reported negative extreme weather event impacts using clustering. In our second stage of analysis we then apply the resulting extreme weather event impact clusters, alongside our measure of average negative event impacts, to examine differences across sociodemographic characteristics using bivariate and multivariate analysis approaches. Clustering methods are frequently used to group survey respondents based on similarities in question responses [70], including grouping respondents with similar levels of reported experience with adverse events [71]. In the context of this research, we apply clustering to identify groups of respondents with a composite of similar extreme weather event impact experiences. To do so, we apply the k-means clustering algorithm [72] to an input of the six extreme weather event impact question responses and select the number of cluster centers using visual heuristics as guidance. To provide further interpretation of these clusters, we examine average extreme weather event impacts across respondents within each cluster and assign cluster names using their dominant event impacts. We choose a naming convention where dominant events impacts are identified when average event impacts within each cluster round to moderately impacted or higher. ‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬ After event impact clusters are identified, we use these clusters to analyze differences across sociodemographic characteristics in our second stage of analysis. These include chi-square tests for bivariate analysis, as well as multilevel modeling for multivariate analysis using the R package lme4 [73]. We use multilevel modeling to account for the spatial grouping of respondents within geographic areas [74], as different areas in the United States have differing levels of impacts from extreme weather event types. We conduct both binary logistic regression analysis (when the dependent variable is dichotomous) and linear multilevel regression analysis (when the dependent variable is continuous). For modeling impact clusters derived from k-means clustering obtained in the first stage of our analysis, we use a binary logistic regression where each impact cluster is a dependent variable, sociodemographic characteristics are modeled as first-level predictors (i.e., fixed effects) and Census Divisions are modeled second-level predictors as random effect intercepts. These first-level predictors include Black, Hispanic, Below Federal poverty threshold, Bachelor’s degree or higher, Female, and Non–metropolitan area. Using these same first- and second-level predictors, we next model the mean composite index of reported extreme event impacts across all six event types as a dependent variable, using linear multilevel regression to accommodate the continuous dependent variable. We also test multiple alternative model specifications, reported in Tables D, I in S1 Text. Using linear multilevel regression, we predict impacts from single extreme event types in separate models (Tables E–G in S1 Text). We also test various household income measures (Tables H, I in S1 Text) as well as model Census Divisions as fixed effects instead of random effects (Table D in S1 Text). We find that our main findings are robust to these alternative model specifications.

4. Discussion We find that a large proportion of the American public self-reported being negatively impacted by extreme weather events in the past three years. An overwhelming majority (86%) of our representative sample of the U.S. public reported being at least slightly impacted by an extreme event and one-third (34%) reported being either very or extremely impacted by one or more extreme event. These extreme weather experience estimates are higher than other national polling conducted during approximately the same time period [26], although a direct comparison cannot be made because we asked about experience over a longer time horizon (3 years vs. 1 year), the level of negative impact experienced, and multiple event types. Moreover, our findings could be capturing very recent trends in extreme weather experience. In 2021, more than 40 percent of Americans were living in counties impacted by climate disasters and more than 80 percent experienced a heat wave [76]. We also found that self-reports of impacts from extreme weather events are related to region, with respondents from certain areas of the U.S. being more likely to report impacts from certain event types, notably extreme heat and wildfire impacts in the West, cold events in the South, and extreme rainfall / flooding in the Midwest, consistent with extreme events exposure in these regions. However, after controlling for sociodemographic factors in modeling and using measures of event impacts that allow for grouping respondents into impact clusters or as an average of extreme event impacts, we did not find geographic location to be the primary driver of reported extreme weather impacts among our participants. When examining the association between sociodemographic characteristics and self-reported extreme event impacts, we find that respondents living in households with income below the poverty level, as well as respondents who are Black or Hispanic, reported more severe negative impacts from extreme weather events, leading us to accept our first three hypotheses (accept H1, H2, & H3). We did not, however, find that respondent age related to reported impacts from extreme weather events (reject H4). We also find that female respondents and those with lower educational attainment reported more negative extreme weather impacts in some of our analysis, leading us to partially accept these hypotheses (partially accept H5 & H6) as these results are not consistent across all analytical approaches. We find similar results for our respondents from rural areas: we partially accept our hypothesis (partially accept H7) as the lowest impact cluster was more likely to have metropolitan residents and we also uncovered a relationship—albeit weaker than poverty and race/ethnicity—to living in a non-metropolitan area and higher average levels of reported extreme weather impacts. Our findings suggest that many people in the U.S. feel they are being negatively impacted by extreme weather events, though these perceptions differ across subgroups. Most prominently, people of color and the poor–groups identified as being more vulnerable to extreme events and disasters in previous literature [10, 27]–reported more negative impacts. Our research only considers self-reported impacts from extreme weather events and does not make any assessment about the objectivity of these reports or seek to identify vulnerable populations. Additionally, we do not consider some other potential demographic factors related to vulnerability, such as respondents with disabilities [77], who might also report being disproportionately impacted and should be a focus of future research. While we asked respondents about negative impacts from extreme weather events within the past three years, there is likely a recency bias in their responses, with recent event experiences being more likely to be recalled [78]. Indeed, our survey was administered during hurricane season, only a few weeks after Hurricane Ida impacted the Gulf and East Coasts, the fifth costliest tropical cyclone in terms of damages in the history of United States [79]. However, there were a record 20 separate weather/climate disaster events with losses that exceeded $1 billion dollars in 2021 [80], including a winter storm in the deep south and Texas, wildfires across the western U.S., and tornado outbreaks in December after our survey was administered. Thus, any time period in which national surveying is conducted could be influenced by recent events. Furthermore, we did not collect more specific information about self-reported negative extreme weather impacts, such as if they are related to health, property damage, lost income, etc., some of which may have different levels of impact severity for respondents with different sociodemographic characteristics. We know that, for example, in the wake of Hurricane Sandy, different population groups were impacted in different ways [81], so measuring these impacts with more specificity in future studies could help better understand these dimensions. Indeed, there is a tradeoff in our approach. While we were able to ask questions relevant to a national sample about multiple different types of events, we were unable to target specific geographies with more specific questions about extreme event experience. This makes it challenging to relate negative reported impacts to named events. Longitudinal research designs, using repeated sampling of the same respondent, may be able to overcome some of these challenges [82], but linking perceptions of negative impacts with measures of objective impacts will likely remain an ongoing challenge [25]. Additionally, further work should be done to develop survey measures that include more impact specificity (e.g., type of impact, severity, frequency, etc.) while maintaining generalizability to multiple event types and avoiding respondent fatigue. Our research also suggests that self-reported impacts from extreme weather events could augment objective measures, such as economic loss and health outcomes, which are themselves difficult to accurately assess [83]. Such self-reports could prove particularly important because some weather impacts have been shown to be underreported by the media and government agencies, such as recurrent household flooding [84], and are also found to disproportionately impact the financially insecure and communities of color [85]. In this respect, some impacts from extreme weather that may have been hidden through conventional reporting methods could become clearer through regular national surveys of the general public. That being said, being more specific about what constitutes an extreme event, an on-going area of scholarship [7], could help provide more precision about perceived event impacts, as well as lay the groundwork for developing subjective vs. objective comparisons. However, such efforts should not invalidate the reported experiences of those impacted by events, as they likely represent real hardships and loss. Indeed, those who have more cumulative experience with hardship and loss, especially among the most vulnerable, may even have lower levels of self-reported impacts from events compared to those with less historical experience and greater means of recovering from them. Therefore, understanding the factors related to these differentially reported impacts should be an on-going effort of both the academic and policy communities, especially with respect to understanding their relationship to objective impact measures. Climate change is already changing the frequency and severity of extreme weather events and will continue to do so in the future without comprehensive global efforts to reduce carbon emissions. While there is substantial uncertainty about population exposure to future extreme weather events, unusual or abnormal weather events may become more common, exposing populations to weather impacts that they may not have much prior experience with or preparation for (e.g., smoke from Western wildfires impacting populations in Midwest and Northeastern states) [86]. Moreover, future impacts could be even further exacerbated by the risks posed by compound events, both weather and non-weather related [12, 87]. These realities suggest that we could be reaching an inflection point where regular national-level surveying of the U.S. public about their perceived experiences with extreme weather may be a necessary component of gauging population impacts and providing information to assist in policy responses. And while disaster preparedness and climate change adaptation planning may be different across regions, planning efforts should account for how different populations may be disproportionately impacted given the new extreme weather threats posed by climate change.

Acknowledgments The authors would like to thank our survey respondents and NORC at the University of Chicago for generously allowing us to include survey questions on their AmeriSpeak Omnibus panel.

[END]
---
[1] Url: https://journals.plos.org/climate/article?id=10.1371/journal.pclm.0000026

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.