Extreme Weather Events and Climate Attitudes, with Lawrence Rothenberg

Abstract: Extreme weather events are the primary way in which most humans experience climate change (US Global Change Research Program, 2014). We study whether witnessing climate change itself — through experiencing changes in extreme weather events (EWEs) — during the 2012-2022 period increased people’s support for climate change mitigating policies. We find that the American public did not respond or respond only selectively to EWEs, and that response is observed only in more recent years. Specifically, before 2019, EWEs did not seem to matter. Since 2019, however, of the four groups we study – Democrats and Republicans with and without college education – only college-educated Democrats increased their support for mitigation policies after experiencing positive shocks in damage from high-temperature events (in particular, droughts and wildfires), and they responded to shocks from these events only. That hurricanes and floods account for 94% of damage to a typical US county from climate change related EWEs but no group responded to them suggests that political actors seeking to increase public support for climate change mitigation policies should emphasize the link between these events and climate change. Additionally, contrary to some previous research, we find that shocks in some events can have effects lasting for five years.

Effect of Primary Election Rules on Politicians’ Rhetorical Extremism

Abstract: When primary elections are closed, in the months leading up to primary elections, candidates for political office may use ideologically more extreme rhetoric to compete for the votes of their co-partisan voters. Previous studies of the relationship between primary election rule and politicians’ rhetorical extremism have examined only one congressional election, thus not controlling for the political environment, or focused on only two presidential elections and thereby limited by the number of candidates. This research improves the literature by studying more than 14 years of complete data and using a causal inference framework. Using large language models and a differences-in-differences approach, I exploit changes in some states’ primary election laws to study if laws mandating more inclusive primary elections reduced the level of extremism in US House Members’ tweets. Of eight cases of legal changes, I find at-best weak evidence of moderation in only one case, and no evidence consistent with moderation in the remaining seven cases.

Climate Change Adaptation Policy Making and Politicians’ Climate Attitudes

Abstract: About half of Congressional Republicans have in recent years expressed doubt about climate change science. What explains politicians’ climate attitude? This research studies the role of local climate change adaptation policy making on politicians’ view about climate change. Using complete sets of congressional tweets and medium-and-large cities' directly elected mayors’ tweets created between 2012 and 2023, several trained large language models, and two different measures of city governments’ flood adaptation effort, I find that city governments' flood adaptation effort, regardless of which of the two different ways in which it is measured, reduced nearby cities' mayors’ support for climate change mitigation policies, an evidence that is consistent with an income effect. Local governments’ flood adaptation effort reduced Congressmen’s expressed doubt about climate change, but it does not seem to have an effect on other aspects of their climate attitude. Additionally, this research contributes to the literature on federalism by providing widespread evidence that politicians' expressed view on an issue can be greatly associated with the office that they hold. Specifically, when the analysis is restricted to overlapping purely urban, or also urban-suburban, or also dense suburban congressional districts and cities, a congressman is, in every case, more than 140 times as likely as a mayor to express doubt about climate change science, but only 2-4 times as likely to express belief in it.