Measuring Self-Preferencing on Digital Platforms

Digital platforms use recommendations to facilitate exchanges between platform actors, such as trade between buyers and sellers. Aiming to protect consumers and guarantee fair competition on platforms, legislators increasingly require that recommendations on market-dominating platforms be free from self-preferencing. That is, platforms that also act as sellers (e.g., Amazon) or information providers (e.g., Google) must not prefer their own offers over comparable third-party offers. Yet, successful enforcement of self-preferencing bans -- to the potential benefit of consumers and third-party actors -- requires defining and measuring self-preferencing across a platform. In the context of recommendations through search results, this research contributes by i) conceptualizing a "recommendation" as an offer's level of search engine visibility across an entire platform (instead of its position in specific search queries, as in previous research); ii) discussing two tests for self-preferencing, and iii) implementing them in two empirical studies across three international Amazon marketplaces. Contrary to consumer expectations and emerging literature, our analysis finds almost no evidence for self-preferencing. A survey reveals that even if Amazon were proven to engage in self-preferencing, most consumers would not change their shopping behavior on the platform -- highlighting Amazon's significant market power and suggesting the need for robust protections for sellers and consumers.