How Not To Bid On Ebay

[eric47]

The study says it may pay to snipe, not that it always does.

 

And in this case the devil may be in the details. If you read the Q&As at the bottom, the study eliminated auctions with two or less bidders. Including them may have changed the result. An auction with only one bidder was obviously won on the first bid, whether sniped or not. A two person bid auction can also be won on the first bid if the first bidder's proxy exceeds the second bidder's bid.

 

Actually, right the devil may be in the details, what makes you think the study eliminated auctions with one bidder?

 

The Q&A are with two people who did *not* make the study but who have done research on the same topic.

 

eBay must have looked that way to the authors of this article, since they report that in one day they have data from 264,073 auctions involving 384,058 distinct bidders. On the other hand, when I look at those numbers, what strikes me is that there were fewer than 2 bidders per auction in their data. To put it another way, a lot of auctions in their dataset had only a single bidder. Obviously conclusions about sniping are going to be different in such auctions (and in our analyses we normally exclude them). [Alvin Roth, my emphasis]

 

Roth isn't part of the Korean study (that would be Byungnam Kahng and Inchang Yang); here's another reference to the article. Roth says specifically that the "a lot of auctions in their dataset (of the Korean study) had only a single bidder", but Roth and his colleague normally exclude them. So insofar as the Korean study took as a dataset not only one but also two bidders, their conclusions are based on those as well.