The next session at the 2026 International Communication Association conference in Cape Town starts with Rongxin Ouyang, who will introduce us to a new social media measure called Bartik instrument. We have for many years discussed the bidirectional relationship between audience engagement and news production: news media set agendas for audiences, but news engagement by audiences also affects how news media select which topics they cover. The causal nature, direction, and strength of this relationship remains unresolved.
This study explores this relationship for Facebook and Twitter: it gathered data from CrowdTangle (2.2m posts) and the Twitter Academic API (4.5m posts) for the period between January 2015 and December 2019. It draws on the Bartik instrument, which was used in the past for examining the relationship between local and national employment rates.
The application of this instrument shows that both pathway directions are supported, but results differ between Facebook and Twitter. There is significant asymmetry on Facebook, but not on Twitter. There is also topic-level heterogeneity as well as outlet-level heterogeneity, but this may be driven by a handful of ideological outliers. Patterns also differ over time, especially in the context of the 2016 US presidential election.
