In the produce auctions held in Vineland, New Jersey, the auctioneer must give the winning bidder the slip of paper the winner needs to claim the lot. He does this by putting the slip into an old tennis ball with a slit in it and throwing the ball to the winner. This saves several seconds per lot for the approximately fifty buyers and fifty waiting farmers. The time needed to select and document the winner of an auction is an example of a transaction cost. Auction theory largely explains a world without transaction costs. This is now changing, and the change is having dramatic effects on auction theory.
First of all, as the introductory example indicates, participating in the auction may be costly. Transaction costs can be a factor in choice of auction form. For example, the Dutch auctions used in the Dutch wholesale flower markets take less than one fifth of the time of the English auctions used in the Vineland auctions. When we teach auction theory, we often mention this--- but it is absent in the models. In addition, continued participation in an auction can have costs that affect bids and choice of auction form.
Secondly, learning his value is not free for a bidder. Indeed, in some cases cost estimating or value estimating is a major undertaking. Some bidders may choose to scrimp on those expenditures and bid cautiously, and others may want the very best estimate they can get. Still others may decline to incur the cost of preparing a bid since the expected profit from participating in the auction is less than the bid preparation cost. Thus, a theory of auctions with transaction costs can be a theory of auction entry. It has recently been shown that when bid preparation costs vary continuously with the quality of the value estimate, there are no reasonable dominant strategy auctions. Of similar importance is the development of a theory of the use of sniping in hard-close auctions like those on eBay to influence the likelihood that competitors will go to the trouble to prepare an aggressive bid.
Third, bid takers incur costs in preparing documentation on what is to be auctioned, in defining lots, and in deciding whom to qualify as bidders. They must balance the value of getting additional bidders against the cost of qualifying them, and also take account of the possibility that their decision to qualify additional bidders may affect the assumptions and behavior of previously qualified bidders.
Fourth, there are costs of publicizing auctions and gathering bidders. These can affect the timing of auctions and lead to grouping of auctions for similar items. Once several items are being auctioned at approximately the same time, the possibility of using combinatorial auctions arises. This raises many new issues, the least studied of which involve transaction costs. The winner determination problem is NP complete and thus potentially costly to solve. There are important issues related to the bidders' costs of preparing and communicating bids: if bids on all possible combinations are required, as in some auction forms, this cost could grow exponentially with the number of items being auctioned. Sophisticated preference elicitation methods may be needed. Even the cost of transferring ownership rises: it is easier to say that X has won the entire state of California than to record that he has won a particular group of unconnected counties.
Auction theory has been an exciting area of research this past ten years because of the many complications that arise from asymmetric information and the way different auction rules aggregate that information. Auction theories dealing with transaction costs are now being developed, however, and this conference will bring together researchers working on them. We hope that not only will participants hear each other ideas, but that their interaction will generate new questions and new insights about which auction forms are superior for which kinds of situations.
Next: Call for Participation
Workshop Index
DIMACS Homepage