When Partisans Fly: Twitter Community Notes and the Political Economy of Social Media Disinformation

Abstract

As concerns over political disinformation grow increasingly prevalent on social media platforms, crowdsourced content moderation schemes have emerged as a possible method of addressing these concerns. Twitter's Community Notes is the first social media platform to implement such a scheme, and in this study, I examine the effects of Community Notes on engagement with political Tweets. Using a novel dataset collected via public data releases and the Twitter API as well as a model of Bayesian media consumption, I find evidence that Community Notes decreases engagement at a significant level at-large, but conditioning on Tweets by politicians, the effect disappears. In addition, consistent with strategic models of media consumption, I find that Democratic agents tend to be sensitive to the partisan bias of the source cited in a Community Note, while Republican agents tend to reject Community Notes overall. This suggests that from a social planning perspective, Community Notes and similar crowdsourced content moderation interventions may be successful in moderating speech at low cost and with less controversy than other stronger forms of content moderation.