Measuring Self-Preferencing on Digital Platforms

Digital platforms facilitate exchanges between platform actors, such as trading between buyers and sellers. However, providers of digital platforms also compete with other actors, referred to as third parties, on their own platforms. In such settings, self-preferencing can occur if the platform treats its own offerings better than comparable third-party offerings—a practice often deemed inappropriate. However, detecting self-preferencing is challenging. This article addresses this challenge conceptually and empirically by putting forward a conceptual framework that defines self-preferencing and conceptualizes two self-preferencing tests. It implements this framework empirically in two studies across three international Amazon marketplaces using a novel metric to measure a product's nonpersonalized visibility. The aggregate findings provide little evidence for self-preferencing in both studies. However, the more disaggregated findings at a country and product category level vary from weak self-preferencing to strong self-depreferencing. As implementing self-preferencing tests requires researchers to choose between several empirical alternatives and assumptions, the approach proposed herein includes extensive sensitivity, specification curve, and heterogeneity analyses whose results support the robustness of the findings.