Tom Cunningham. (@testingham) First version Jan 31 2023, data last updated April 2023, text updated July 2024.1
- 1 Thanks to comments from Sahar Massachi, Katie Harbath, Nichole Sessego, David Harris, Theodora Skeadas, Eric Davis, Jimin Lee, and many others.
This note describes the suspension practices of the major social media platforms. I have collected a dataset of around 200 suspensions of prominent people across 12 platforms between 2011 and early 2023, stored in a google spreadsheet. The chart below summarizes the full dataset:
The data helps illuminate what platforms are doing. It is very difficult for an outside observer to see how a platform moderates their content. The advantages of studying the suspension of prominent users are that (1) the data is public and (2) the outcomes are comparable across platforms.
Key findings.
- The rate of suspensions has grown over time. The increase seems to be primarily due to adoption of new policies rather than changes in user behaviour or changes in enforcement.
- Suspension practices are fairly similar across the major platforms. Meta, Twitter, and YouTube all have broadly similar policies: they each suspend users for hate speech, election misinformation, COVID misinformation, and incitement.
- The most common reasons for suspension were hate speech (15%) and COVID misinformation (12%). Platforms typically do not publicly state the reason why they suspend an account, however I was able to code the majority of cases either because the reason was clear from context, or the platform’s justification was relayed to the suspended user or a journalist. In 19% of cases I could find no reason given. For cases with a reason there were quite a wide variety: hate speech (15%), covid misinformation (12%), incitement (7%), and posting personal information (6%).
- Twitter suspended more people than other platforms. From examining cases it seems this was primarily due to differences in the type of content posted rather than differences in policies or in enforcement.
- US politicians were suspended at a much higher rate than non-US politicians. This was concentrated on Twitter and seems to be a mixture of US politicians being more active, being more likely to make policy-violating statements, and being under more scrutiny.
- Among US Federal politicians suspended, 8 were Republicans, none were Democrats. The Republicans were suspended for a variety of different reasons and on a variety of platforms. The asymmetry does not seem to me to be primarily a difference in enforcement but a higher propensity for Republicans to say or do policy-violating things.
I am working on a separate essay about why platforms suspend users. It is difficult to give clear reasons why platforms suspend users. In a separate essay I try to break down how much their action can be attributed to influence from owners, from employees, from users, from advertisers, or from governments. Having this dataset of suspensions is very useful to be able to make generalizations about platform behavior.
There are three prominent peaks in the history of suspensions:
Reasons for suspension over time
It is clear that Twitter has suspended more people than Meta and YouTube, but the majority of suspensions fall in categories which are also enforced by Meta and YouTube.
Between 2015 and 2017 there were a series of alt-right personalities suspended from Twitter.
All Wikipedia-reported Twitter suspension, highlighting accounts with more than 1M followers (not all suspensions list the number of followers).
Suspensions for accounts with >1M followers
