How Do Content Producers Respond to Engagement on Social Media Platforms?

Abstract

When creating new content, many social media users hope to receive engagement from other users. This research examines how receiving that engagement affects different users' subsequent behavior on the platform. We address this question through a field experiment on Twitter in which some users' posts were purposefully shown more often to other users, which (on average) increased the amount of engagement they received. We estimate a doubly robust instrumental variable model that allows us to estimate individual-level treatment effects, and we find substantial heterogeneity across users in terms of how they respond to additional engagement: most users do not significantly change their behavior, but some users respond by substantially increasing their time spent on the platform, posting more content, and engaging more with other users' content. Users who respond most strongly are systematically different than the rest of the user base on observable pre-experiment user metrics, thereby providing substantive insights about which users value engagement very highly. Our results demonstrate how social media platforms can increase content creation, content consumption, and overall usage of the platform by focusing on this group of users and targeting them with interventions that are intended to increase the amount of engagement they receive.

Keywords: Two-sided platforms, social media, customer engagement, causal machine learning

JEL Classification: M31, C26, C31