Sensing the Pulse of the Pandemic: Demographic Disparities of Public Sentiment toward Covid-19 through Social Media

Keywords: Covid-19, social media, sentiment analysis, demographic bias
Abstract Type: Paper Abstract

Abstract

The global Covid-19 pandemic has caused severe mental stress to different social groups. Social media offer a unique platform to observe public sentiment toward Covid-19. However, social media data are biased toward younger, well-educated, and wealthier populations. Sentiment analysis based on Twitter data without considering the underlying demographic biases may overlook the stress of certain social groups and lead to unfair estimations. Therefore, this study analyzed the Twitter data in the U.S. in 2020 and 2021 to answer two research questions: (1) What are the geographical and demographic disparities of public sentiment toward Covid-19 reflected on social media? (2) How can we alleviate the demographic bias within social media to fairly evaluate public sentiment toward Covid-19? To address these research questions, we developed a social media mining framework for measuring sentiment toward Covid-19 and constructed a Sentiment adjusted by demographics (SAD) Index considering social media user demographics for unbiased sentiment measurement. The results showed that demographic disparities of sentiment toward Covid-19 existed, and female Twitter users and users under 18 years old were more mentally sensitive to the dreadful effects of the pandemic in general. The SAD index corrected the overestimation of negativity in 30 states, especially in Vermont, and the underestimation of negativity in 21 states. After the adjustment by demographics, the most negative and positive states were Wyoming and Vermont, respectively. The knowledge will shed light on Covid-19 perceptions among diverse social groups in each phase and inform governments to customize pandemic mitigation strategies for different communities.