IOP@FSU Field-Advancing Research: Do Twitter Bans Matter? Political Influencers and the Quality of Information they Share Before and After Account Suspensions

The Institute of Politics at Florida State University (IOP@FSU) is a world-class, nonpartisan, and nationally renowned institute that promotes engagement in politics by students and citizens. Housed within the College of Social Sciences and Public Policy, the IOP@FSU supports applied political research by a cadre of world-class scholars and mobilizes the talents of our alumni, students, faculty, and friends all in the heart of Florida’s Capital City. This research was funded by the IOP@FSU. Learn more about IOP@FSU here.           

Increasing media attention has been paid to account bans on Twitter. Beginning with the banning of Donald Trump’s Twitter ban for content that incited violence on January 6th, 2021, some conservatives have accused Twitter of biased censorship (Bovard 2021; Porter 2021). With the rise and fall of Elon Musk’s purchasing of Twitter, tabloids buzzed with the potential for Musk to return previously banned accounts to their owners, including Trump’s (Milmo 2022; Siddiqui, Harwell, and Dawsey 2022). Although Twitter account bans occupy much attention nationally, research on the actual effects of account bans is scarce. Ultimately, this creates a context which is being framed as an infringement on free speech while those “infringements” are not evaluated by empirical research.

By observing tweets pertaining to the Arizona Audit before and after the banning of 8 high-profile accounts, this research project addresses this gap of knowledge. During Donald Trump’s Big Lie that the 2020 general election was fraudulent, he and portions of the Republican Party endorsed a highly controversial and partisan audit of Maricopa County, Arizona’s election (Levine and Massoglia 2021). During this audit, Twitter banned 8 high-profile accounts responsible for spreading misinformation. Using an original sample of over 245,000 tweets preceding and following this ban, this project explores any changes in the quality and bias of information being shared as well as attitudes towards the audit itself. Ultimately, our findings suggest that, in the case of the account bans surrounding the Arizona Audit, very little changed in the quality and bias of information shared or the activity of political influencers.


            On July 27th, Twitter banned 8 accounts pertaining to the Arizona Audit citing platform manipulation and spam. The bans included the official @ArizonaAudit account, an independently run account @AuditWarRoom, and six spinoff accounts. Tweets which included the keywords “Arizona Audit” were collected from July 15th to August 6th using DiscoverText. These dates surround the account bans by 9 days on either side.

            This yielded a sample of over 245,000 tweets. The “exact duplicate” function of DiscoverText was used to sort tweets into groups and identify political influencers, which were operationalized as accounts which posted tweets receiving ≥ 200 retweets. A total of 148 tweets reached this criterion. The number of retweets these tweets received were highly positively skewed, 75% of tweets received less than 951 retweets while the maximum number of retweets received was 23,831. Tweets in this final sample disproportionately appeared before the ban; of those tweets which received ≥ 200 retweets, 85 appeared before the ban, 18 appeared on the day of, and 45 appeared after the ban.

            This project performed a content analysis on our tweets and the accounts they belonged to in order to answer these three questions: Does banning accounts on Twitter change who shapes the conversation around the Arizona Audit? Does banning accounts on twitter change the kinds of arguments political influencers make about the Arizona audit? Finally, does banning accounts on Twitter affect the type and quality of information shared by political influencers?


            Table 1 describes the biography descriptions of accounts appearing in our sample of tweets with ≥ 200 retweets before, the day of, and after the ban. For preservation of space, only the most pertinent codes were retained, and all others collapsed into “Other”.  The “Fourth Branch” category includes public-facing news outlets. Notably, tweets belonging to these outlets were more likely to appear after the ban compared to tweets coming from other accounts and tweets belonging to accounts with right-leaning political sentiments were more likely to appear before the ban compared other accounts. The increase in tweets belonging to news outlets after the ban could suggest that the ban itself made the Audit more “news-worthy” and attracted media attention. Tweets belonging to accounts with right-leaning political sentiments in their biography description appearing significantly more in the period before the ban compared to tweets belonging to other accounts signals a dip in the presence of conservative accounts discussing the Arizona Audit receiving more than 200 retweets.

            Table 2 describes the relationship between tweets supporting or opposing the audit before, the day of, and after the ban. Tweets which oppose the Arizona Audit, which presumptively entails endorsing the official results of the 2020 general election, are significantly more likely to appear after the ban than before the ban. However, tweets supporting the audit were not significantly related to timing. This suggests an influx of tweets opposing the audit after the ban, but a persistence of those who support the audit. This finding seems to partially compliment and contradict the findings from Table 1. Although it makes sense that increased news outlet attention to the audit could accompany increased opposition to the audit, the persistence of support for the audit contradicts assumptions that a dip in tweets belonging to conservative accounts would accompany a dip in tweets supporting the Arizona Audit.

            Finally, Table 3 concerns the relationship between quality and bias of information shared and the timing in relation to the account bans. Here the lack of significant relationship between Partisan Left sources and timing as well as Partisan Right sources and timing is most pertinent. Partisan Left/Right was coded when Ad Fontes Media Bias Chart (2022) rated news sources as lower quality and higher bias. That tweets sharing Partisan Left/Right sources are not significantly related to timing suggests the ban did little to affect the sharing of this lower quality and higher bias information. Also notable is the persistence of Mainstream sources. Mainstream sources were those rated by Ad Fontes Media to have higher accuracy and lower bias, and its lack of relationship with timing suggests that the ban did little to encourage the sharing of these higher quality sources.

            Cumulatively our findings create a mixed, yet pessimistic view of the effect of Twitter bans on political influencers and the quality and bias of information they share. This conclusion is based on the decrease in conservative-identifying accounts receiving ≥ 200 retweets coupled with the increase in news outlet coverage and tweets opposing the Arizona Audit. These findings would suggest that the ban introduced Audit-opposing audiences to the discourse, however the persistence of Partisan Left/Right sources signal that the quality and bias of information did not significantly improve. Nor were higher quality sources of information more likely to appear after the ban. In the face of platform polarization (Conover et al. 2011) and cries of censorship from conservative news outlets (Lanum 2022), it may be time to reconsider whether banning accounts is the most effective way to encourage civil discourse and reliable information.

Kyle Rose is a PhD student in the Department of Sociology at Florida State University.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.