Working Paper: CEPR ID: DP17260
Authors: Martin Peitz; Anton Sobolev
Abstract: Biased recommendations arise naturally in a market with heterogeneous consumers: a seller offers a product to a mix of consumers who can purchase through an intermediary or directly from a seller. "Picky" consumers are uncertain about match quality, which they observe only after purchase, while "flexible" consumers are always happy with the match. Therefore, picky consumers rely on the intermediary's recommendation. We provide conditions under which the intermediary will recommend a welfare-reducing bad match with positive probability, resulting in inflated recommendations. Regulatory interventions may lead to higher social welfare. However, a regulatory intervention that prohibits recommending bad matches may backfire.
Keywords: intermediation; digital platforms; recommendation bias; recommender system; asymmetric information; experience good; ecommerce
JEL Codes: L12; L15; D21; D42; M37
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Intermediary's profit motives (L21) | Inflated recommendations (E31) |
| Inflated recommendations (E31) | Recommendation of bad matches to picky consumers (L15) |
| Intermediary's inability to commit to recommendation policy (D52) | Recommendations that do not match consumer preferences (D11) |
| Regulatory interventions (G18) | Higher social welfare (D69) |
| Certain regulations may backfire (L51) | Lower social welfare (D69) |