|
Article Excerpt
I. INTRODUCTION
The answer to the question of how bidders choose to enter one auction versus an alternative competing auction is important to understand in providing a complete picture of the nature of endogenous entry decisions into auctions. The issue of how a bidder chooses which among alternative auctions to enter as well as whether to enter an auction at all is of substantial importance to auctioneers attempting to attract bidders to their auctions. There is a growing literature addressing issues in endogenous entry in auctions. There are several papers such as Engelbrecht-Wiggans (1993), Pevnitskaya (2004), and Palfrey and Pevnitskaya (2008) that consider the decision of whether to enter an auction or not and study the effect of endogenous entry on bidding behavior. There is a much larger group, for example, Harstad (1990), Engelbrecht-Wiggans (1993), Levin and Smith (1994), McAfee (1993), and McAfee and McMillan (1987), that considers competition between auctioneers in which the auctioneers are competing along some characteristic such as entry price. We consider the case in which auctioneers may be competing with each other based on the choice of auction mechanism. Our interest is in the degree to which the auction mechanism itself might impact the entry decisions of potential bidders. Of course, there would be a direct pecuniary impact of the mechanism in that bidders should be more likely to enter auction mechanisms that they believe allow them to get lower prices, but there may be other cognitive issues in regard to the structure of the mechanism that may also impact the choice behavior.
In Ivanova-Stenzel and Salmon (ISS, 2004a), we introduced an experimental design aimed at eliciting and measuring preferences subjects might possess for different auction institutions and applied the design to testing bidder preferences between ascending (a dynamic mechanism) and sealed-bid first-price (a static mechanism) auctions. It seems reasonable to expect bidders to have preferences between these two auction mechanisms due to the strategic differences between them. In a first-price auction, the winner pays a price equal to his own bid, whereas in the ascending auction, he pays a price that is an increment above the second highest bid. These differences might result in difficulty of deciding how to bid and in the possibility of experiencing some form of regret. (1)
The design involved having subjects make a choice between entering into a two-bidder sealed-bid first-price auction versus a two-bidder ascending auction, each with different entry prices. By varying the entry prices, we could measure the subjects' willingness to pay for either auction. Our main finding in the paper was that while subjects exhibited strong preferences for the ascending auction at equal entry prices, they were not willing to pay up to the expected profit difference between the two auctions to participate in an ascending auction. Thus, they ended up entering the supposedly less preferred sealed-bid first-price auction even in situations in which they would expect to make a higher net surplus (including the cost of the entry fee) by entering into the ascending auction instead. That paper proposed one possible answer to the puzzle that involved assuming that bidders were risk averse in their bidding behavior and then risk averse to the same degree in their auction choice behavior. This explanation worked because neither risk-averse nor approximately risk-neutral bidders should be expected to pay much to enter into an ascending auction instead of a sealed-bid first-price auction just as we observed in the experiment. For the most risk-averse bidders, this is because while they bid the highest in the first-price auction and therefore expected little surplus in the event that they win, they consequently end up winning with higher probability than less risk-averse bidders. This is a trade-off risk-a-verse bidders are willing to accept, and consequently, it is easy to show that highly risk-averse bidders should not be willing to pay much to get into the more lucrative, though risky, ascending auction. At the other end of the range, bidders who were less risk averse bid lower in the first-price auction, which means that they expect a larger surplus when they win but it also means that the expected surplus differential between the two formats is not so large. These less risk-averse bidders will pay a little to get into the ascending auction because of their lower probability of winning the first-price auction but not much.
While the hypothesis of risk aversion was able to match with the data, applying risk aversion to explaining bidding in auction data is controversial. There is a long-running debate among auction theorists and experimental economists running back to Cox, Roberson, and Smith (1982), Cox, Smith, and Walker (1988), and Harrison (1989) regarding whether risk aversion is a reasonable explanation for the bidding behavior in first-price auctions. These papers led to a long round of comment articles starting with the December 1992 issue of American Economic Review as well as many other related papers. Many of the key points in this debate are summarized in Kagel (1995). Because of the accumulation of so much doubt regarding whether the fundamental bidding behavior itself can be well characterized by risk aversion, there is certainly reason to suspect that the compound hypothesis that risk aversion explains both the bidding behavior and the auction choice behavior may ultimately be unsatisfying. This leads us to consider alternative explanations.
A. Loss Aversion
The most plausible alternative explanation is loss aversion. This model of decision-making behavior is suggested quite forcefully in Rabin and Thaler (2001) and Rabin (2000) as being superior to risk aversion (at least when losses are possible). In ISS, losses were guaranteed for the losing bidder in the auction because of the fact that they had to pay the entry fee. Even the winning bidder could lose money in a round if they failed to earn enough to offset the cost of entry. If bidders are loss averse, then they would be unwilling to pay a higher entry fee for an auction even if the expected value for that auction is higher taking the entry fee into account. This behavior is consistent with the results observed in ISS, and the experiment could not disentangle this explanation from the risk aversion explanation.
B. Structure of the Bidding Process
Another reasonable alternative explanation has to do with the difference in bidding structure between the sealed-bid first-price (static) and the ascending (dynamic) mechanisms. In the first-price auction, a bidder sends in a bid and they are done immediately. In the ascending auction, bidders have to wait for a while as the price clock ticks up, which may take a more substantial amount of time. The additional time also requires the subject to actively pay attention to the screen for a longer period of time. Prior lab and field experiments show evidence that bidders in ascending auctions demonstrate nonnegligible impatience or perhaps more accurately described, a desire to have the auctions end faster to reduce the amount of time and effort on the bidding process. For example, Lucking-Reiley (1999) finds that bidders in a field experiment often requested that he use higher bid increments in his auctions to push the auctions along faster. Shachat and Swarthout (2002) finds in a lab experiment that sellers in an ascending procurement auction that proceeded in a fairly slow manner drop out of the auction earlier than they should, leading the buyers to pay more than they should have to. The authors state that "We conjecture that the tediousness of the English auction is responsible for the early exit behavior." Evidence confirming that the behavior in both situations is consistent with impatience is found in Isaac, Salmon, and Zillante (2005), which presents a test of a model of bidding in ascending auctions that incorporates impatience and finds that it fits the data better than the competing alternatives. (2) In these last two sets of experiments, subjects were willing to give up nonnegligible earnings to have the auctions concluded faster. This was true even though subjects went through the series of auctions as a group, and one auction closing faster did not do much to speed the overall experiment session because it went forward at the speed of the slowest auction each round.
Based on these findings, it could well have been the case that what was observed in ISS was due to a manifestation...
|
