Online Auctions: Strategies, Reputations, and Trickery
Though auctions may require lots of preparation when done in-person, they are incredibly easy to perform over the web. On online platforms such as eBay (the most popular of them all), auctions are easy to set up and run. Even though these auctions are designed to resemble second-price sealed-bid auctions, do these online auctions play out the way people expect? Since records of these auctions are kept within a database, people can analyze this data to examine the behavior of the auction participants. As it turns out, the fact that online auctions occur over the internet may play a crucial role in determining how bidders bid and whether they even enter.
The form of auction used by eBay implements a “proxy bid,” in which bidders specify their maximum (ceiling) bid, and allow another machine to increment their starting bid until either the machine reaches the ceiling specified by the bidder or the bidder becomes the current highest bidder. Of course, bidders may submit a new proxy bid if they wish to stay in the auction. This generally continues until the auction’s time limit ends. This may at first seem to be a reasonable way to carry out auctions remotely via the internet, but this methodology has a few shortcomings we will arrive at very soon.
In a paper from the Journal of Economics by Patrick Bajari and Ali Hortacsu, the authors have observed some patterns in the way people bid in eBay’s auctions. Firstly, the fact that the auction is limited by time provides another variable for participants to play around with: how far into an auction should bidders submit their bid? Bidding early means that the bidder reveals his private “personal value” early on in the auction. Intuitively, this gives the rest of the bidders a small advantage, since the other bidders can respond with bids that slightly exceed said value.
In fact, this is consistent with what Bajari and Hortacsu observed: data shows that many participants employ a strategy called “sniping,” in which participants submit their bid right before the auction ends. This is also sensible since if a bidder ends up with the highest bid right before the auction ends, the other bidders do not have the time to “react” to the new bid by submitting a new, higher bid. At the same time, however, bidders need not reveal their actual value early; they can just as easily bid lower than their personal value until the end of the auction, when they would submit a higher bid. As such, the process of bidding involves complicated strategies that somewhat resemble those used in second-price sealed-bid auctions but with additional quirks.
Another perhaps more obvious consequence of auctions using the internet as a medium is the concern that bidders may develop regarding the transactions that will occur. How safe is it to participate in online auctions? While we may be relatively sure that eBay takes some security precautions, people may ask whether certain auctioneers (the sellers themselves) are trustworthy people. There in fact exists a way to approximate (this may be misleading in some cases we will discuss later) their trustworthiness: auctioneers’ reputations can be altered through the use of a “feedback system,” in which bidders report their auction experiences (positive or negative). This can help people avoid auctioneers who seem suspicious and interact with those that are not.
In a paper by Zoonky Lee, Il Im, and Sang Jun Lee, the authors observe that this kind of feedback system actually may influence the bids people make. When participating in an auction whose auctioneer has a large amount of negative feedback, the bidders will give a lower minimum bid than those with more positive feedback (this can be observed by studying auctions of similar products). This sounds reasonable, since bidders may feel a certain “risk” when the auctioneer has a poor reputation, and account for this by putting less money on the line. However, to what degree this trust affects people’s bids appears to depend not only on the level of trust or distrust, but also on the condition of the product that is being sold. The authors of the paper suggest that when the auctioneer is selling a brand-new item, negative feedback is less impactful, whereas for “used” items, the bidders are more cautious and reserved when deciding to enter the auction. So, it appears that many online bidders already have some intuition of what makes a trustworthy auctioneer.
But there is still some online deceit that makes it through people’s outer defenses. Auction fraud has become one of the most common kinds of “Internet crime” reported by users, and there is frighteningly no systematic, effective way to prevent this. The fact that there is no such method is unsurprising, since fraudsters are not uniform in their approaches and are difficult to identify, but the result is that online bidders are required to be vigilant and cautious about making transactions. These online fraudsters can also easily take advantage of the feedback systems described previously, so feedback systems are not a reliable way to detect fraud. This will be a challenge that eBay and others will need to address in coming years if they wish to keep online auctions a safe experience for all.
In our Networks course, we have restricted our study of auctions to several rules and several well-known types of auctions. For example, we assume that people cannot submit multiple bids, and that people win an auction by bidding the highest (or in English auctions, being the last bidder remaining). A brief examination of online auctions shows that these rules do not necessarily apply. However, we still see elements of second-price sealed-bid auctions in play, and there are still certain “equilibria” we can observe from data. As mentioned earlier, bidding right at the end of an auction seems to be a promising strategy for bidders; bidding earlier only gives the other bidders hints regarding what your true “personal value” may be. While online auctions may be more complex in nature than the examples we have studied, the large amount of data saved from online auctions allows people to study them closely. The last link I have posted below describes a possible systematic method of detecting auction fraud using a network analysis algorithm (NetProbe); I look forward to seeing how people will approach this issue.
References:
http://www.jstor.org/stable/pdf/1593721.pdf?acceptTC=true
http://dl.acm.org/citation.cfm?id=359753
http://repository.cmu.edu/cgi/viewcontent.cgi?article=1530&context=compsci