The 2020 Presidential Election on Twitter: An Exploration of Candidates’ Social Presence, Campaign Momentum, and the Effect of Misinformation

Abstract

Twitter has evolved from a site of inconsequential information spread to an instant primary source used as the preferred outlet to discuss and witness any semblance of news the emerges each day. Political outreach thrives in this medium due to ease of communication across a diverse and engaged audience. The rapid nature of Twitter creates an ecosystem suited for misinformation spread. Influential misinformation afflicted the platform during the 2016 U.S. presidential election and during the COVID-19 pandemic, causing Twitter to implement policy changes that, while minor, garnered significant media attention and user backlash.

This thesis aims to understand the characteristics of political Twitter as it relates to prominent campaigns, and how misinformation affects campaign momentum. Specifically, we investigate the 2020 presidential election between Donald Trump and Joseph Biden through tweets authored by the candidates and through tweets mentioning the candidates. We evaluate the candidates' social presence on Twitter, their campaign momentum, and the effect of misinformation via sentiment analysis, network analysis, and momentum analysis. In terms of social presence, we find that Trump's tweeting behavior is more active and that the tweeting behavior of both candidates increases in the days surrounding a notable event. We also observe that Biden more frequently publishes emotionally charged messages of trust, anticipation, and joy. Both Biden and Trump exhibit a negative campaign style when mentioning their opponent, but only Trump's negative style resonates with his base as seen via a significant increase in user engagement following Trump's negative tweets. In discussion on Twitter, we discover that general election information is more commonly tied to Biden while non-election information and niche stories are more often tied to Trump.

Regarding campaign momentum, we find that Biden's campaign witnesses more positive momentum relative to Trump's campaign as measured by a sentiment indicator. We then apply sentiment and curvature indicators to explore associations between targeted misinformation events and changes in campaign momentum. Overall, this thesis demonstrates the use of Twitter as a data source to investigate the social presence and momentum shifts of influential members.