Abstract
This paper studies the incentives of a subscription-funded platform that offers both proprietary and third-party content to bias its recommendations about which con tent users should consume. Consistent with Netflix’s practice, we consider fixed-fee bargaining between the platform and a content provider, which eliminates any static incentive to bias recommendations. However, our dynamic model identifies two dis tinct incentives to bias recommendations: improving the platform’s future bargain ing position and increasing users’ expected surplus. The former favors first-party content, while the latter favors the ex ante superior content. As a result, biased recommendations may lead to either self-preferencing or third-party preferencing.
Keywords
Recommendation, Platform, Algorithm, Signal Jamming;
JEL codes
- D83: Search • Learning • Information and Knowledge • Communication • Belief
- L42: Vertical Restraints • Resale Price Maintenance • Quantity Discounts
Reference
Doh-Shin Jeon, and Mikhail Drugov, “Dynamic Recommendation Bias”, TSE Working Paper, n. 26-1742, April 2026.
See also
Published in
TSE Working Paper, n. 26-1742, April 2026