Applied Micro Seminar - Tiffany Tsai (NUS)
Title: Steering via Algorithmic Recommendations
Abstract: We study a platform's incentive to maximize the value of data in their recommendation. We ask if and how Amazon's dual identity as an information intermediary and a seller may affect its data-driven ``Frequently Bought Together'' recommendations (FBT). We document that Amazon-selling products receive 70% more FBTs while sending a similar number as non-Amazon-selling ones. We show that (1) controlling price and sales, the same product receives 8% fewer FBTs during Amazon's temporary absence; (2) steering is stronger in categories where FBT is estimated to be more effective; (3) recommending non-Amazon-selling recipients is estimated to be more than 50% efficient than recommending Amazon-selling recipients.