Dissecting Netflix's Self-Preferencing: Evidence from Viewer-Level Data

Self-preferencing by dominant digital platforms has become a focal point for antitrust scrutiny, yet little empirical work has examined this behavior in the context of video streaming. This paper provides the first systematic analysis of self-preferencing on a subscription-based streaming platform, focusing on Netflix. We assemble a novel dataset that combines a weekly panel of Netflix’s U.S. catalog from 2016 to 2025, official Top 10 rankings since 2021, Wikipedia page views as an external proxy for popularity, and device-level streaming data from tens of millions of U.S. smart TVs. We begin by showing that the exit of licensed series significantly increases the likelihood of subscriber churn, whereas the effect of movie exits is small and even slightly negative. This underscores the risks of dependence on non-original serialized content and motivates Netflix’s incentives to promote Originals. We then document that Netflix Originals are substantially more likely to appear in the Top 10 rankings than non-originals, conditional on popularity and availability. The magnitude of this self-preferencing effect is comparable to the influence of popularity itself, especially for serialized content. Finally, using a difference-in-differences design with matched titles, we show that Top 10 inclusion has a significant causal impact on subsequent viewer engagement, with stronger effects for Originals. Taken together, our findings suggest that Netflix leverages interface prominence to steer attention toward its proprietary content while insulating itself from the risks associated with expiring licenses, raising important implications for content competition and platform governance in the streaming era.