Impact of Extreme Weather on Public Sentiment: Empirical Evidence from China

• In Event: Climate Affects Sentiment; Sentiment Affects Outcomes: Culture, Attitudes, and Environmental Policy

Abstract

Climate change has significantly increased the frequency, duration, and intensity of extreme weather events across the globe, presenting not only environmental and infrastructural challenges but also profound social and psychological consequences. While a growing body of literature has documented the physical damages and economic disruptions caused by these events, their psychological impacts on collective well-being and public sentiment remain underexplored—particularly in the Global South and non-Western societies (Campbell-Lendrum et al., 2023; Lawrance et al., 2022; Newman & Noy, 2023). This study aims to fill this gap by examining how extreme weather events influence public sentiment through the lens of civic expression in China, a country undergoing rapid climate adaptation and digital civic engagement.

We analyze citizen-generated content from the Message Board for Leaders (MBL), a prominent national e-participation platform that enables individuals to voice their concerns to government officials directly. Drawing on a longitudinal dataset of 3,084,890 messages collected between 2011 and 2023, we employ a multi-method analytical framework to capture the dynamic relationship between climate anomalies and public sentiment. Specifically, we assess the impact of extreme weather events—including temperature anomalies, heavy precipitation, and high-velocity wind episodes—on the emotional tone of citizen messages submitted during and after these events. To address the challenges of sentiment measurement in large-scale, user-generated Chinese text, we construct a manually annotated subsample of over 50,000 messages as training data. We then fine-tune large language models (LLMs) to classify the sentiment of the whole dataset, enabling scalable and context-sensitive estimation. To enhance robustness and cross-validate findings, we also implement lexicon-based sentiment analysis methods designed for Chinese syntax and semantics (Chen et al., 2019). Integrating machine learning and rule-based approaches ensures methodological rigor and interpretability.

Our results reveal a robust association between extreme weather events and negative sentiment expressions in civic discourse. The strength of this relationship varies by event type, with temperature anomalies and heavy rainfall eliciting the strongest adverse reactions. Local climate resilience infrastructure, such as drainage systems, emergency response capacity, and climate-sensitive urban planning, is a significant moderating factor. Cities with stronger resilience measures experience significantly less sentiment volatility during adverse weather periods. Further stratified analyses uncover substantial heterogeneity in these effects across cities with different economic development levels, geographic zones, and degrees of urbanization.

Theoretically, our study contributes to emerging literature at the intersection of climate change, political psychology, and public administration by introducing “online participatory sentiment” as a novel behavioral indicator of collective emotional response to environmental stressors. This indicator provides real-time, granular insights into how populations emotionally interpret and react to climate events, complementing traditional survey-based approaches. Practically, our findings have several implications. First, they highlight the importance of incorporating emotional resilience and psychological well-being into climate adaptation planning. Second, they demonstrate the potential of participatory governance platforms like the MBL as low-cost, real-time early warning systems for detecting public distress. Third, they call attention to the uneven emotional burdens of climate change, reinforcing the need for equity-focused climate governance that addresses physical vulnerabilities and mental and emotional dimensions.

Authors

• Zhengsheng YU, Hong Kong University of Science and Technology (Guangzhou)
  Presenting Author

• Ruoyun Wang, The Hong Kong University of Science and Technology(Guangzhou)
  Non-Presenting Co-Author

• Tianle Liu, Harbin Institute of Technology (Shenzhen)
  Non-Presenting Co-Author