The Declining Price Anomaly in Dutch Dutch Rose Auctions

In symmetric, independent, private-value auctions of a single object, with risk-neutral bidders, the English (second-price) auction has strategic simplicity. It is optimal for bidders to reveal their valuation of the object and make bids accordingly. In that case, the auction is won by the person with the highest valuation who pays a price equal to the second-highest valuation. In a Dutch (first-price) auction, this simplicity vanishes. Here, winning bidders have to pay their bid. To make a profit, they shade their bids and bid below their valuation of the object. Somewhat loosely, one may state that an English auction is truth-revealing whereas a Dutch auction requires strategic behavior. The simple structure of the one-unit English auction vanishes if two identical objects are auctioned sequentially. Now, in the first round, it is optimal for bidders to shade their bids to account for the option value of participating in the subsequent second round (Robert J. Weber, 1983). Bidders with a higher valuation also have a higher option value. Therefore, they shade their bids in the first round by a greater amount than do bidders with a lower valuation. As the auction proceeds, the number of bidders decreases. Over the sequence of auctions, the number of objects decreases as well. The first fact has a negative effect on the competition for an object and the second has a positive effect. Both effects cancel out and prices follow a martingale. As a result, all gains to waiting are arbitraged away and the expected prices in both rounds are the same. The latter result also holds for sequential auctions of more than two objects and does not depend on whether the auction is English or Dutch. This neat theoretical result is not supported by empirical research, which usually finds price declines. Orley Ashenfelter (1989) found a mild price decrease in sequential auctions of identical units of wine. McAfee and Daniel Vincent (1993) also present empirical evidence on sequential wine auctions. They found that, on average, the second unit of wine was sold at a price 1.4 percent lower than the price of the first unit. Victor Ginsburgh (1998) uses data from wine auctions in which the auctioneer acted as an agent for bidders who were not present at the auction. He found price declines but, apparently, the absent bidders entered bids that did not fit with the theory. Empirical studies by Ashenfelter and David Genesove (1992) (see also Jean-Jacques Laffont, 1997) and Alan Beggs and Kathryn Graddy (1997) concern heterogeneous objects. They also detected price declines. Because of the contradiction between theory and empirical studies, the declining price phenomenon is regarded to be an anomaly.