Partisan and Geographic Variation in Emotional Responses to COVID-19 Vaccination on Social Media

Public acceptance of the first COVID-19 vaccine administration was not uniform, yet the nature and county-level characteristics of heterogeneous emotional responses remain poorly characterized.

To examine changes in social media emotional expression across US counties before and after the first COVID-19 vaccine administration, and whether changes varied by county-level partisanship and COVID-19 death rates.

Cross-sectional study of geotagged COVID-19-related posts on the socia media platform Twitter from 3065 US counties from over 1.9 million unique users who created more than 18 million posts during the study period (September 1, 2020, to March 31, 2021), covering 100 days before and after first vaccine administration. Data were analyzed from January 2023 to April 2026.

Geotagged COVID-19-related social media posts during the 100-day periods before and after December 14, 2020.

Levels of joy, sadness, anger, and fear expressed in posts, measured using the National Research Council of Canada Word-Emotion Association Lexicon.

After vaccine administration, fear decreased (β = -0.424; 95% CI, -0.510 to -0.338; P < .001; Cohen d = -0.20), joy increased (β = 0.683; 95% CI, 0.601 to 0.766; P < .001; Cohen d = 0.33), and anger increased (β = 0.445; 95% CI, 0.384 to 0.506; P < .001; Cohen d = 0.28). Sadness did not change significantly (β = -0.004; 95% CI, -0.075 to 0.067; P = .91). Democratic-leaning counties showed greater joy increases (β = 0.162; 95% CI, 0.115 to 0.209; P < .001), larger fear decreases (β = -0.066; 95% CI, -0.115 to -0.017; P = .007), and smaller anger increases (β = -0.042; 95% CI, -0.075 to -0.009; P = .009) than Republican-leaning counties. Higher county COVID-19 death rates were associated with greater fear decreases (β = -3.69 × 10-6; 95% CI, -6.03 × 10-6 to -1.35 × 10-6; P = .002); death rate did not significantly moderate joy or anger.

In this cross-sectional study of 18 million geotagged social media posts, the first COVID-19 vaccine administration was associated with increased joy and anger and decreased fear, with fear showing the greatest evidence of a vaccine-specific shift. These findings suggest that monitoring social media discourse can provide early signals of optimism, skepticism, and division, thereby informing targeted communication strategies.