eBay as the “Terminator”: Determining User Suspension From Feedback Ratings

Abstract

Good reputations are won and lost by individuals and companies in various forms on the Internet. eBay has the earliest and most well-known formal ecommerce reputation system and “terminates” members who do not play by the rules. We examine a dataset of 102,035 Not-A-Registered-User (NARU) eBay account feedback ratings from eight different countries. Careful investigation of feedback density over time reveals that threshold values for termination can be detected from feedback history. Several modeling approaches are studied and the Local Polynomial Regression Model (LPRM) is chosen for analysis because it is data driven and no prespecified parametric functional form is assumed. Analyzing US data, we find that large-volume sellers have a higher overall negative feedback compared to those of medium volume sellers. We use the second derivative of the LPRM to explicitly determine the threshold points, that is, the point where a seller's negative feedback increases significantly. These thresholds vary from as much as 18.5 weeks prior to termination (for Medium sellers), to as little as 8.5 weeks (for Large sellers). We find similar results in the other countries. These threshold points provide valuable information for buyers when choosing a seller. Our analysis suggests that eBay's termination policy appears to be correlated to fee revenue.

Keywords:

• reputation
• trust
• online auctions
• termination
• feedback

ACKNOWLEDGMENT

This article is based on the thesis work of Mendonca, conducted under the joint supervision of Wang and Hayne. Hayne's research was partially supported by the National Science Foundation (grants SES-9819031, SES-0196032) and Wang's research by NSF grants DMS-0706761, DMS-0854903 and DMS-1106975. The authors are grateful for the contribution of C.A.P. Smith to this project.

Notes

¹This was eBay's policy from 2003–2005. The feedback data that we have used for this project are from the same period.