Abstract
Reputation systems contribute a lot to the success of e-commerce. This study utilizes an institutional feature to make causal inference on the effect of the number of ratings on demand. Platforms like eBay and Taobao assign sellers into different levels based on their rating scores. Therefore, continuous rating scores lead to discontinuity in levels, allowing a regression discontinuity (RD) design. I collected a large and unique dataset with nearly three thousand sellers and 6.8 million transactions on Taobao. Implementing a non-parametric panel RD, I find that on average, having one extra level increases sales by over 9%, otherwise equal. The effect is even stronger for established sellers with rating scores in the range of thousands. It is also highly heterogeneous across industries. I build a simple dynamic pricing model that predicts sellers keep lowering their prices before the thresholds, and increase them after passing the thresholds. The finding on price changes is largely consistent: Sellers keep lowering their prices for up to 2% before reaching thresholds and increasing them afterwards.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call