The Spread of "China Virus" on Twitter
BERT embeddings classify reactions to Trump's inflammatory language
In March, as COVID-19 was undergoing its first uncontrolled spread throughout several US states, President Donald Trump began referring to the disease as the “Chinese Virus”. His tweet was widely commented and shared, and garnered over 150,000 likes. It also triggered scores of political reactions, anti-Asian discrimination, and thousands of subsequent social media posts on the topic.
In this post, we use Natural Language Processing (NLP) to analyze the discourse around the term “Chinese Virus” on Twitter. We find that the conversation was largely generated by President Trump’s use of the term, and that geographic variations in those supporting or criticizing the term largely follow partisan political lines. These results demonstrate that political leaders have a very strong influence on how their followers use language, and point to a deeply polarized country.
One Tweet can go a long way
Donald Trump first used “Chinese Virus” on Twitter on March 16th, claiming that COVID-19 “comes from China” when pressed on the issue. A widespread debate ensued, with many health experts cautioning against the use of the term because of the stigma it might cause and with Democratic politicians decrying it as racist. In the subsequent months, however, expressions such as “Chinese Virus”, “Wuhan Virus”, and even “Kung Flu” have become commonplace in Trump Administration communications, among right-wing commentators, and on social media. Anti-Asian harassment linked to the COVID-19 pandemic has become a global issue, with most attacks placing blame on Asians for causing the virus or spreading it to others. In the US, the Stop AAPI Hate organization recorded over 2,500 incidents of anti-Asian violence from March 19th through August 5th. Inflammatory language that blames Chinese people for the pandemic undoubtedly fueled this violence.
Cuomo wants "all states to be treated the same." But all states aren’t the same. Some are being hit hard by the Chinese Virus, some are being hit practically not at all. New York is a very big "hotspot", West Virginia has, thus far, zero cases. Andrew, keep politics out of it....
— Donald J. Trump (@realDonaldTrump) March 17, 2020
For the purpose of this project, we collect all the tweets related to the term “Chinese Virus” from January 1st, 2020 to April 30th, 2020. Overall, 14,984 posts are included in our analysis. Initial time-series plots show that while the term was already in use within small circles before Donald Trump’s tweet, its use went through the roof in the days following the president’s endorsement. Since we restrict our analysis to geotagged content, we are able to pinpoint where users were located when they reacted to the president’s tweet. The county-level map below shows that reaction and subsequent use of the term “Chinese Virus” took place across the entire country.
As could be expected, much of the social media activity was concentrated in dense urban areas such as Miami, Los Angeles, Houston, and New York. In fact, at state level, prevalence of the topic is mostly proportional to population. Largely, the term “Chinese Virus” was on the mind of social media users everywhere, and number of related posts is surprisingly independent of state-level political leanings and socioeconomic characteristics. One notable exception is the District of Columbia—where the topic was remarkably salient within a small population, likely due to the concentration of news outlets and political users.