Communications - May 2008

Understanding fraud with respect to Internet auctions is especially important because the oft-cited “network externality” in which having large numbers of buyers leads to large numbers of sellers, which leads to even more buyers, and so on. It is based upon the knowledge that the winner of the auction will receive what he or she was expecting. If more than a handful of buyers perceive that the system does not work in a fair and neutral manner, the entire network effect may start unraveling.

If the claim of such a low level of fraud were true, an auction site could encourage a higher level of buyer activity (and prices bid) by charging a minimal surcharge on the final winning bid to provide automatic insurance to buyers. Analyzing one auction site, we found that a one-dollar increase in the final winning bid translates into over $700 million in annual dollar volume. A surcharge of one promille per dollar of the final price above the additional transaction costs such as bookkeeping, investigation of claims, and issuing of compensation should be enough to provide insurance and to leave ample additional profit for the auction site operator. The fact that insurers do not offer such policies calls into question industry’s estimates of fraud incidences.

This procedure is not as simple as it sounds, as having automatic insurance would have an effect on the behavior of both buyers and sellers. Buyers will be willing to tolerate a higher level of risk and bid higher even for unknown sellers, or sellers with lower levels of reputation. Another complication is that it creates incentives for crooked sellers and buyers to form coalitions in which they swindle the auction operator or the insurance entity. The advantage of an auction site operator compared to the occasional buyer or seller is that the operator has the resources and experience to develop methodologies that detect and reduce such swindling operations to a minimal level. By providing automatic insurance, the auction site operator would benefit from the higher level of volume in bidding activity.

We concentrate on buyers being swindled. In our study we discovered that sellers face a similar problem of cheating; however, auction houses have devoted extensive resources toward protecting sellers, but have invested limited effort into protecting buyers. Why? We contend that if sellers do not put items up for auction, the site cannot survive, while bidders/buyers on their own do not justify the existence of an auction site.

Another factor involves the fact that most sellers use the auction site multiple times (in many cases, thousands of times). As a result they develop experience and methods that protect them. Bidders, on the

other hand, have limited experience with the auction process and do not have the expertise to protect themselves. ¹For example, in most auction sites, sellers do not ship a product to the buyer until the buyer has paid for the product.

RESEARCH SUMMARY

We employed a variety of methods to undertake an exploratory investigation of Internet auction fraud. We performed a literature review of prior empirical studies on Internet fraud and auction fraud. This led to a preliminary and exploratory survey conducted in 2003. Our survey results also helped guide our participant-observation exploration: We bought and sold on major auction sites from 2003–2005 to better understand how swindlers work and what actions might be taken by the auction houses and by buyers. We also interviewed bidders and sellers that contacted us through our own buying and selling.

In our exploratory pilot study (see [ 5]), we surveyed 1,298 winners of Internet auctions at a major auction site to see whether they received what they were expecting. The respondents came from 14 different item categories and a full range of prices. We asked primarily two questions and collected other information from the auction site itself.

1. Did you receive any item after you won the auction in question?

2. If you did receive an item, was it what you were expecting?

We interviewed willing respondents to our survey. It was difficult to gather the data using automatic means, so we did everything manually. Generally speaking, auction houses put many obstacles to prevent automatic data gathering for such a research investigation. They limit the number of interactions one can have with other members per day. Thus, it is not obvious how to do such a study.

Staying within the rules set up by the auction houses, we were prevented from using mechanical methods such as Web crawlers; otherwise, it would have been much easier to contact a large number of auction winners. Further, the auction houses are willing to give researchers information, but what they offer is mostly meaningless for such a study. For example, looking simply at feedback does not convey much information. It carries a value of 0, neutral, or 1, so there is no magnitude. For example, a “winner” in Athens, Greece, whom we talked to lost close to