The effect of different extreme weather events on attitudes toward climate change

examine. Some research focuses specifically on extreme droughts [19], evaluates temperature and precipitation anomalies [24], or considers extreme flooding [25]. Other scholars aim to examine a range of weather events that could be associated with climate change [16, 21]. Nota- bly, only some of these papers provide causal evidence [26, 27]. Together, these methodological differences create limitations in the overall understanding of how extreme weather events may influence mass attitudes. Even though this literature remains unsettled regarding whether and how any given extreme weather event might impact social and political opinions, several important trends have emerged. One is that political ideology can play a critical role in how “evidence” of climate change through an extreme weather event is interpreted. In surveys of individuals who have experienced extreme weather events, political ideology is consistently a major factor in climate policy support, which indicates that motivated reasoning is at play [22, 28]. This is also evidenced in discussions following exposure to extreme weather. In a systematic comparative-case analysis of post-disaster communities, the authors found that residents in Democratic communities were more likely to link a given weather event to climate change [23]. Previously held beliefs and other personal characteristics may be key to understanding how and why certain extreme weather events influence opinions [12, 13]. Other important factors may include the intensity of

a given event and the type of policy measure proposed (i.e., mitigation vs. adaptation) [17]. Research design Studying the political effects of disasters within a causal framework is complicated by the ever- present potential of hidden biases. Exposure to an extreme weather event could be correlated with a range of political outcomes, which can confound analysis and undermine key inferences [29]. For example, there is evidence that the experience of a major weather event like a hurri- cane is associated with socioeconomic status [30, 31]. For that reason, many scholars are turn- ing to approaches like a difference-in-differences design to account for potential confounding factors. In this paper, we similarly exploit cross-time and cross-country variation in exposure to extreme weather in the U.S. using a dynamic DiD approach. Data We construct an exposure indicator using Federal Emergency Management (FEMA) data for federally declared disasters that may be associated with climate change [32] (see [21] for an example of the use of FEMA data to identify areas exposed to an extreme weather event). This data provides a summary of major disasters and other emergencies beginning in 1953, includ- ing severe flooding, volcanic eruptions, landslides, and chemical waste exposures. Data is pro- vided at the county-level and includes the start date of each disaster declaration as well as an incident type (e.g., flood, tornado, etc.). We construct a binary indicator of exposure to an extreme weather event associated with climate change one year before capturing the outcome

using panel survey data (years 2010, 2012, and 2014). For example, for the panel wave imple- mented in 2010, we use FEMA data for 2009. We use this approach since our expectation is that the effect of exposure to a disaster on concerns about climate change will take some time to crystallize since disaster victims usually emphasize concerns about their living conditions just after a disaster rather than paying attention to public or collective concerns [33], such as the impact of climate change. Also, this means that there is a gap between the occurrence of the event and the survey; however, this should not undermine the possibility of finding any effects since previous evidence found that disasters can have long-lasting effects [34]. As a robustness check, we used events that happened the same year as the survey but before its PLOS ONE Extreme weather events and attitudes toward climate change PLOS ONE | https://doi.org/10.1371/journal.pone.0300967 May 15, 2024 3 / 10