What are people tweeting about the 2020 Presidential Election?

A Research Report by Jian Cao, Nicholas Adams-Cohen, and R. Michael Alvarez

The public continues to passionately discuss the 2020 Presidential Elections. In a world of high social media participation, not only do citizens consume vast amounts of news before, during, and after election day but can also actively share their views and opinions with others online. This active participation leads to fierce discussions and debates of electoral issues, one of the factors that potentially led to record-breaking turnout in the 2020 national election.

As one of the most popular social media platforms, Twitter continues to play a particularly important platform in this year’s elections. On Twitter, users discuss all aspects of the election, including allegations of voter fraud, voting on election day, potential electoral challenges, and remote voting. The topics discussed on Twitter evolve quickly and often, changing as major events occur and news stories break.

Given this year’s increased attention to election administration and security, monitoring Twitter data can offer important and valuable information about election integrity. This is particularly true in the context of 2020, with the COVID-19 pandemic making it very difficult to conduct in-person election observation. Our group has collected and analyzed Twitter data about election administration and voting technology since 2014, and for this election cycle are posting real-time visualizations of the data on our Monitoring the Elections website.

Our Twitter monitor collects all tweets that mention select election keywords, performs real-time analytics including geolocation, and creates informative visualizations on our election dashboard. We provide three types of analytics: hourly and daily frequencies of election topics, word clouds of the trending voter fraud and election challenge topics, and maps that visualize which states users are from and discuss. Please refer to the white paper for additional details about how Twitter monitors work.

To highlight some of the results from our work, we first display tweet frequencies over time by discussion topic. Among the six topics we track, the majority of discussion prior to and on election day consists of tweets concerning election day voting, voter fraud, and remote voting. After November 3, 2020, election day itself, the discussions shifted towards voter fraud and election challenges.

In the word clouds, we display the most frequently used words in the tweets concerning voter fraud and election challenges every 15 minutes, with the size of the word corresponding to how often the word appears. Some common words such as “voter fraud,” “election fraud,” and “rigged election” remain in the word clouds at all times, while some interesting topics such as “concede nothing” and “release evidence” rise and fall.

The state maps show which states are being discussed in the context of our tracked topics. In the map below, the most frequently mentioned states in the context of voter fraud tweets are shown in darker red. By browsing through the daily maps, we can see how the discussions shift from Nevada and Georgia to Pennsylvania and Michigan.

At this point, we’ve not had a chance to take a deep dive into our collected data to ascertain the extent to which individual users, news, media, and political organizations, or automated accounts (“bots’’) are sending tweets. In our previous work, we looked for evidence of bot activity in past election cycles and found little evidence that bots significantly altered our core findings. You can take a look at the recent book that Adams-Cohen and Alvarez wrote with Silvia Kim and Yimeng Li, Securing American Elections, for further discussion of this point. While certainly possible that the 2020 Presidential election data is similar to our previously collected datasets, the fact that we did not find significant evidence of bot activity in the past doesn’t mean that they are not present in the new data. We will continue closely analyzing the data in the coming weeks and will present updates on this and related topics.

We argue that this crowd-sourcing of election monitoring is an effective way to study election successes and problems in real-time, particularly in the 2020 general election. In addition to the real-time analytics shown above, please stay tuned for further analyses that help you understand how social media and election influence each other.