Living in a warming world – climate impacts on society

Climate change remains one of the greatest challenges of the 21^st century. Humans are not only the key driver of climate change, but are also affected by its consequences, with profound impacts on human life arising through changes in environmental and social circumstances. Climate change impacts on human behaviour are observed via two principle data types which both reflect daily human activity: social media and mobility data. Specifically, social media data are employed to analyse temperature impacts on hate speech online. Even though links between temperature and physical aggression are known, it remains unclear how these patterns extend to online environments, where hate speech can have detrimental consequences for the mental health of the affected persons. Using machine learning classifiers to identify hate speech in four billion geolocated tweets, we show that temperature has strong non-linear impacts on the occurrence of online hate speech across the USA with hate-tweet levels remaining low at moderate temperatures but sharply rising during both hot and cold extremes. This pattern persists across income groups, religious beliefs, and election outcomes and even various climate zones, including those where heat is common which suggests adaptation limits to hot temperatures. A complementary analysis for six European countries finds quasi-quadratic, nonlinear temperature impacts on digital racism and xenophobia across Europe. To assess not only the impacts of heat but also the ability to adapt we employ mobility data from New York City. An analysis of daily passenger data of more than 400 subway stations over six years shows that there is not only a strong, non-linear temperature impact on subway usage but also disparities between neighborhoods with respect to the capacity for heat mitigation. Correlations between neighbourhood-level mobility reductions and socioeconomic indicators suggest that the ability to reduce mobility on hot days is afforded by those that also hold other privileges, hence leading to unequal, compounding health impacts in disadvantaged neighbourhoods. Finally, we harness a combination of mobility data, Google search trends and Covid-19 data to explore how behavioural responses may develop under a prolonged or repeated risk exposure. The econometric approach explores changes in the response to Covid-19 risk through two channels: risk perception and the resulting behavioural response, i.e. mobility reduction. Across both channels, the risk response diminishes over time even before the availability of vaccines. This highlights the attenuation of behavioural responses to prolonged risks, with implications for managing long-term crises such as increasingly repeated exposure to weather extremes under ongoing climate change.