Mapping Public Sentiment: An Analysis of COVID-19’s Impact on Emotional Expression in Chinese Cities
The COVID-19 pandemic triggered a wave of unprecedented social and emotional responses across the globe. This study delves into the intricate dynamics of public emotional expression during the initial stages of the pandemic in China, leveraging the power of social media data to provide a nuanced understanding of how sentiments spread and interacted across cities. Utilizing a vast dataset of 45 million geocoded tweets from Sina Weibo, China’s largest microblog platform, the research analyzes public negative emotional expression concerning COVID-19 (PNEEC) across 335 Chinese cities between January 1 and March 31, 2020. This period, encompassing the initial outbreak and subsequent containment efforts, offers a crucial window into the evolution of public sentiment during a rapidly unfolding crisis.
The study quantifies PNEEC through a lexicon-based semantic analysis approach. Researchers employed the Chinese Emotional Vocabulary Ontology (CEVO) and Hownet lexicons to classify the polarity of words in Weibo posts related to COVID-19. By aggregating the sentiment scores of these posts for each city and day, the researchers constructed a daily PNEEC measure. This method allows for a granular analysis of sentiment variations across time and location, offering a more comprehensive picture than traditional survey-based methods. The choice of Weibo as the data source provides a rich and real-time reflection of public discourse, capturing the immediate emotional responses to evolving events. The validation of findings using both CEVO and Hownet reinforces the robustness of the sentiment analysis approach.
To understand how emotions propagate across geographical boundaries and social networks, the study analyzes the interplay of three types of inter-city ties: geographic proximity, mobility patterns, and social media connections. Geographic ties were defined based on shared borders between cities. Mobility ties were derived from Tencent Location Big Data, capturing billions of trips between cities from 2015 to 2017. Social media ties were constructed from a directed inter-city attention network, built from real-time posts on various social media platforms. This multi-faceted approach recognizes that emotional contagion can occur through various channels, reflecting the complex web of interactions in modern society. The use of historical mobility data, while not perfectly reflecting pandemic-era movement, still provides valuable insights into pre-existing connections between cities.
The research further divides the study period into three distinct phases, mirroring the evolving nature of the pandemic: the warning phase (January 1-20), marked by initial reports and rising concern; the isolation phase (January 21-February 29), characterized by strict lockdown measures and reduced mobility; and the normalization phase (March 1-31), where containment efforts began to yield results and life gradually returned to normalcy. This phased approach allows for a more nuanced understanding of how emotional interactions shifted in response to changing circumstances and policy interventions. Analyzing each phase separately acknowledges that factors influencing emotional spread may vary depending on the severity of the outbreak and the prevailing social context.
Recognizing that public emotions are influenced by a multitude of factors beyond social interactions, the study incorporates a comprehensive set of control variables. These include epidemiological data (cumulative confirmed cases, deaths, and recoveries), policy interventions (travel bans, emergency declarations, isolation measures), environmental conditions (PM2.5 levels, temperature, rainfall), and socio-economic indicators (GDP per capita, disaster experience, hospital bed capacity, demographics, internet access, and insurance coverage). This rigorous approach aims to isolate the specific effects of inter-city ties on PNEEC while accounting for other potentially confounding factors. The inclusion of such a wide range of control variables strengthens the validity of the study’s findings.
The study employs several statistical methods to analyze the spatial and temporal patterns of PNEEC. Global Moran’s I is used to measure the overall spatial autocorrelation of emotional expression, revealing how similar sentiments clustered geographically. Local Moran’s I identifies localized emotional correlations between a city and its neighbors, highlighting areas of high or low emotional interaction. Regression discontinuity design assesses the causal impact of policy changes, specifically examining how shifts between pandemic phases affected emotional connectivity. Spatial autoregressive models investigate whether a city’s PNEEC is influenced by the sentiment of its network neighbors, capturing potential emotional contagion effects. Finally, Panel Vector Autoregression (PVAR) and Forecast Error Variance Decomposition (FEVD) are employed to disentangle the contributions of geographic, mobility, and social media ties to local PNEEC, quantifying the relative importance of each channel in shaping emotional responses. The combination of these diverse analytical techniques allows for a robust examination of the complex interplay between social connections and emotional expression. This detailed methodological approach provides a strong foundation for understanding the dynamics of emotional contagion during a major public health crisis, offering valuable insights for future pandemic preparedness and response strategies.