The Celebrity Factor: Exploring the Impact of Influencers on COVID-19 Vaccine Sentiment through Bayesian Modeling of Time Series

Abstract

Online social networks allow for information to rapidly propagate throughout the world, and opinions expressed on such platforms can influence people’s decisions. During the COVID-19 pandemic, many influential public figures used these social networks to share their opinions about the vaccines developed to combat the virus. Many influencers encouraged vaccination, and a considerable number also expressed doubt and skepticism over the efficacy of the vaccines. This study modeled the impact that eleven influencers’ statements had on the overall sentiment towards COVID-19 vaccines, as expressed on Twitter. Sentiment is measured by collecting a series of publicly-available tweets made regarding the vaccine during the pandemic, and assigning each a sentiment score based on the VADER lexicon. Several models were used to analyze the impact of the influencers’ statements, including linear, sequential and tree-based models. The results were obtained by constructing a Bayesian structural time series model based on each model’s counterfactual estimate. The results found that influencers who share messages encouraging vaccination generally tend to increase the number of ”pro-vaccination” tweets over the next 20 days. Influencers sharing ”anti-vaccination” messages sometimes resulted in a decrease in anti-vaccine tweets, and other times in an increase over the next 20 days. The results from this study provide an introductory look into the complex issue of vaccine hesitancy and the effect of influencers on vaccine messaging, and inform public health strategy regarding this issue.