Bye-box: An Analysis of Non-Promotion on the Amazon Marketplace 03.06.2022

Matthias Hunold (1) , Ulrich Laitenberger (2, 3, 4) , Guillaume Thébaudin (2, 3)

1 Universität Siegen [Siegen]
2 ECO-Télécom Paris - Equipe Eco Economie
3 SES - Département Sciences Economiques et Sociales
4 CRED - Centre de Recherche en Economie et Droit

We study seller and product recommendations of the hybrid e-commerce platform Amazon. Using web-scraped data, we find that Amazon makes the visibility of offers of third-party suppliers in the "buybox" dependent on prices on competing marketplaces like Walmart and eBay. Amazon's own offers are visible regardless of their competitiveness. We find that the absence of seller recommendations makes recommendations to related products more effective and Amazon tends to steer consumers in these situations more often to products it sells itself. We discuss that this behavior is difficult to reconcile with the hypothesis of an independent marketplace operator.

Keywords: Amazon marketplace buybox self-preferencing algorithm bias recommendation algorithms D40 L42 L81; Amazon marketplace; buybox; self-preferencing; algorithm bias; recommendation algorithms D40; L42; L81

Domains: Quantitative Finance [q-fin]

Matthias Hunold, Ulrich Laitenberger, Guillaume Thébaudin. Bye-box: An Analysis of Non-Promotion on the Amazon Marketplace 03.06.2022: CRED WORKING PAPER n°2022-4. 2022. ⟨hal-04104183⟩