A theoretical and empirical investigation of multi-item on-line auctions
Information Technology and Management
| Hi-index | 0.00 |
One variant of the Dutch auction roughly corresponds to a multi-unit, progressively ascending, uniform price, lowest winning bid, open auction. The overall revenue realized by the auctioneer in such an auction is given by R = Q 脳 pw where, Q is the number of units of the item being auctioned, and pw is the lowest winning bid price at the end of the auction. R is only dependent on pw (Q being fixed) and an interesting question is whether it is possible to increase R.Proposed in this paper is a method for increasing pw (and consequently R). Given that the maximum bid that a bidder places reflects his or her valuation, our method relies on increasing the valuation of those bidders whose valuations are in close proximity to the lowest winning bid price. We propose achieving this objective through the use of a coupon, which is introduced at an algorithmically determined time during the auction, and which, if introduced, assumes a non-decreasing face value. The coupon can be used by all participants -- the price paid by participants with winning bids utilizing the coupon is discounted by the face value of the coupon at the end of the auction process. Assuming other things are equal, participants with bids at or near the lowest winning bid price during the auction are more likely to use the coupon and increase their bids by an amount in excess of their valuation (the excess being recovered through the use of the coupon) thereby creating additional competition and potentially increasing pw.We present an algorithm that determines, based on present auction dynamics, if a coupon should be introduced, the initial face value of it, and possible subsequent revisions of the coupon face value. We also present a simple example to illustrate the algorithmic steps and illustrate the increase in R that results as a consequence of the proposed methodology.