How COVID-19 has Impacted American Attitudes Toward China: A Study on Twitter

TL;DR

This paper investigates how the COVID-19 pandemic altered American attitudes toward China using a vast Twitter dataset to exploit the outbreak as a natural experiment. It applies regression discontinuity and moving-window difference-in-differences to identify a causal effect of COVID-19 awareness (proxied by tweeting about COVID-19) on sentiment toward China, finding significant declines after exposure (e.g., $Δ_{RD} = -5.28$, $Δ_{DID} = -4.21$). The results show a rapid, pandemic-driven shift to more negative views of China, alongside a surge in China-related tweeting, and a shift in the composition of users toward more negative attitudes. The findings illuminate how self-interest and a behavioral immune response can shape foreign-policy opinions during crises, with implications for public opinion and policy under global health shocks.

Abstract

Past research has studied social determinants of attitudes toward foreign countries. Confounded by potential endogeneity biases due to unobserved factors or reverse causality, the causal impact of these factors on public opinion is usually difficult to establish. Using social media data, we leverage the suddenness of the COVID-19 pandemic to examine whether a major global event has causally changed American views of another country. We collate a database of more than 297 million posts on the social media platform Twitter about China or COVID-19 up to June 2020, and we treat tweeting about COVID-19 as a proxy for individual awareness of COVID-19. Using regression discontinuity and difference-in-difference estimation, we find that awareness of COVID-19 causes a sharp rise in anti-China attitudes. Our work has implications for understanding how self-interest affects policy preference and how Americans view migrant communities.

Figures (8)

• Figure 1: American attitudes toward China declined to a 3-year low of -38 after the outbreak of COVID-19. Attitude toward China is measured by averaging sentiment over Twitter users on tweets that mention China but not COVID-19 on a 7-day sliding window. Sudden and steep declines in China-related sentiment occur nationwide. The declines are accompanied by increases in the volume of tweets on China denoted in the figure with light and dark blue shaded areas. The daily number of users mentioning China on Twitter increased from 5,096 in late December 2019 to $66,535$ by the middle of March 2020. On March 16, over $95\%$ of Twitter users who mentioned China also mentioned COVID-19. See Figure \ref{['SI:102-overall-sentiment-volume']} for a full version of this image.
• Figure 2: Changes in the composition of Twitter users who mention China. Users grouped by sentiment toward China before the COVID-19 outbreak (October - December 2019) are displayed on the left side of the figure. $70.56\%$ of users had negative attitudes, $19.44\%$ had neutral attitudes, and $10.00\%$ had positive attitudes. User distribution over sentiment groups after the COVID-19 outbreak (March - May 2020) are on the right side of the figure. $76.91\%$ users were in the negative group, $15.20\%$ were in the neutral group, and $7.90\%$ were in the positive group. A total of $18.38\%$ of users moved to a more negative group (e.g., neutral to negative) while $11.49\%$ moved to a more positive group. $2.18\%$ users moved from the negative group to the positive group, and $3.74\%$ users moved from the positive group to the negative group.
• Figure 3: Identifying causality with regression discontinuity. The average sentiment of China-related tweets before and after the date a Twitter user first posts a tweet mentioning COVID-19. Individuals are segmented by treatment week, denoted in vertical gray lines.
• Figure 4: Identifying causality with difference in difference estimation.The sentiment of Twitter users toward China declines suddenly in the treatment week, i.e. the week after they post their first tweets about COVID-19 (solid lines). In contrast, the counterfactual group, who had not yet been treated by a given treatment week, maintain stable sentiment toward China (dotted lines). The treatment effect is estimated by first taking the difference in sentiment on China before and after the treatment within groups and then taking the difference of these within-group differences between groups. This suggests that tweeting about COVID-19 causes an immediate decline of $4.21$ in sentiment toward China.
• Figure S1: A full version of American attitudes toward China from January 2017 to June 2020. Attitude toward China is measured as an average over Twitter users on tweets mentioning China but not COVID-19 on a 60-day sliding smooth window. It declines slowly before nosediving suddenly after the outbreak of COVID-19 (solid). The decline is accompanied by an increase in tweet volume on China (light and dark blue areas). Dashed lines represent state-level average attitude toward China in the 10 states with the most Twitter users: California, New York, Texas, Florida, District of Columbia, Kansas, Illinois, Washington, Georgia, and Massachusetts.