Working Paper: CEPR ID: DP17006
Authors: Konrad O. Stahl; Tobias Klein; Xiang Hui
Abstract: Online ratings play an important role in many markets. However, how fast they can reveal seller types remains unclear. To study this question, we propose a new model in which a buyer learns about the seller’s type from previous ratings and her own experience and rates the seller if she learns enough. We derive two testable implications and verify them using administrative data from eBay. We also show that alternative explanations are unlikely to explain the observed patterns. After having validated the model in that way, we calibrate it to eBay data to quantify the speed of learning. We find that ratings can be very informative. After 25 transactions, the likelihood of correctly predicting the seller type is above 95 percent.
Keywords: online markets; rating; reputation
JEL Codes: D83; L12; L13; L81
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| seller's transaction history (N81) | likelihood of buyer leaving a rating (L81) |
| negative rating (Y70) | likelihood of future buyers leaving a rating (D16) |
| negative experiences (D91) | learning about seller quality (L15) |
| positive experiences (I31) | learning about seller quality (L15) |
| prior beliefs about seller quality (L15) | posterior beliefs after transaction (G41) |
| absolute difference between prior and posterior beliefs (D80) | decision to rate (D79) |