Abstract
INTRODUCTION: The COVID-19 pandemic caused a widespread public health and financial crisis. The rapid vaccine development generated extensive discussions in both mainstream and social media, sparking optimism in the global financial markets. This study aims to explore the key themes from mainstream media's coverage of COVID-19 vaccines on Facebook and examine how public interactions and responses on Facebook to mainstream media's posts are associated with daily stock prices and trade volume of major vaccine manufacturers. METHODS: We obtained mainstream media's coverage of COVID-19 vaccines and major vaccine manufacturers on Facebook from CrowdTangle, a public insights tool owned and operated by Facebook, as well as the corresponding trade volume and daily closing prices from January 2020 to December 2021. Structural topic modelling was used to analyze social media posts while regression analysis was conducted to determine the impact of Facebook reactions on stock prices and trade volume. RESULTS: 10 diverse topics ranging from vaccine trials and their politicization (note: check that we use American spelling throughout), to stock market discussions were found to evolve over the pandemic. Although Facebook reactions were not consistently associated with vaccine manufacturers' stock prices, 'Haha' and 'Angry' reactions showed the strongest association with stock price fluctuations. In comparison, social media reactions had little observable impact on trading volume. DISCUSSION: Topics generated reflect both actual events during vaccine development as well as its political and economic impact. The topics generated in this study reflect both the actual events surrounding vaccine development and its broader political and economic impact. While we anticipated a stronger correlation, our findings suggest a limited relationship between emotional reactions on Facebook and vaccine manufacturers' stock prices and trading volume. We also discussed potential technical enhancements for future studies, including the integration of large language models.