Fraud on the Internet is developing into a major issue for consumers, businesses, and governments [1, 6, 10]. The Financial Times in 2003 called online fraud "an epidemic of huge and rapidly growing proportions" and noted the incidence of fraud was 20 times higher online than offline . The complaints of online fraud registered at the IC3 Web sitewhich is jointly sponsored by the U.S. Federal Bureau of Investigation and the U.S. National White Collar Crime Centerhave grown from around 20,000 in 2000 to around 200,000 in 2007, which represents a compound annual growth rate of 39% . At the same time, the dollar value of losses has skyrocketed at an annual rate of 50% from less than $18 million in 2001 to over $200 million in 2007 .
One area that is particularly interesting is auctions. Auction fraud reported to the Federal Trade Commission has likewise grown tremendously from 106 in 1997 to around 24,000 in 2007  and continues to climb dramatically. Internet auction fraud is one of the main sources of overall Internet fraud, with estimates of incidence from 64% to 87% of all Internet fraud [4, 7]. However, major Internet auction sites estimate that fraud is involved only once in approximately 10,000 auctions , which appears to contradict these observations.
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.1 For example, in most auction sites, sellers do not ship a product to the buyer until the buyer has paid for the product.
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 20032005 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 ), 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.
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 $10,000 and felt very strongly about his loss. His situation and (negative) feedback for the seller carried the same weight as someone who lost one dollar in an auction.
Auction houses have devoted extensive resources toward protecting sellers, but have invested limited effort into protecting buyers. Why?
The results of our preliminary survey are as follows: We received 98 responses from the auction "winners," 21 of which either did not receive any item at all, or did not receive what they were expecting (mostly because the item was damaged or otherwise in worse shape than described). Further, eight out of 98 received absolutely nothing at all, which represents 0.62% minimum (assuming everyone else, including non-respondents, received something, as we will discuss). This incidence would be 62 times higher than official estimates of fraud, and this is clearly a subset of all fraud.
If either the buyer received nothing or did not receive what was expected, we call that "swindled" (recognizing, of course, that the buyer may or may not have been intentionally swindled and that this is definitely a superset of purely fraudulent activity). Still, knowing the incidence of "swindling" is important because it most certainly is an input into buyers' perceptions of the fairness of auctions.
We calculate the worst-case rate of negative response (by dividing the total negatives by total responses) and the best-case rate (by dividing the total negatives by total contacted). It depends on how one views the representativeness of our sample to know which is more "accurate." If one feels that the respondents are representative of the population at large, then the worst-case estimate would be closer. If one feels that our respondents are "disgruntled," then the best-case estimate would be closer.
In addition, we have two opposing factors. For those who are embarrassed about being cheated, they will be underrepresented. They do not want to report that they were cheated because it would make them look foolish. On the other hand, there are people who were cheated on one auction and are so mad they want to tell their story, even if we asked them about a different auction, and that will lead to overestimation. The true rate would most likely fall somewhere in between the best- and worst-case estimates.
The percentage of negative responses varied for different categories of auctions (see Figure 1). Computers and electronics had the highest worst-case rate of negative responses: three-quarters of respondents either did not receive their computer or it arrived damaged. In terms of the best-case rates, the worst category was Jewelry (excluding watches), in which 5% of all auction winners contacted either did not receive anything at all or did not receive what they were expecting. Four categories had no negative responses (Paintings; Guitars; Event Tickets; and Watches). One surprise was that Jewelry, excluding watches, had one of the worst rates of dissatisfaction, while Watches had one of the best. In one category, Sports memorabilia, we received no responses at all. Of course, these category measures should be taken with a grain of salt, as the number of responses within each category is quite low. Still, the overall response indicates a worst-case estimate of 21.4% negative, with a best-case (and conservative) estimate of 1.6% negative.
We also found the following contingencies:
Perhaps one lesson to be learned from this breakdown is that buyers appear to be more careful in situations in which the dangers are more obvious.
Based on a survey of literature (for example, [2, 3, 10]), plus our interviews, plus our own participation in auctions, we derived some generalities in the swindling area. First, there are the methods used to actually execute the swindle. Second, there are methods used to avoid appearing fraudulent. The main methods used to actually execute the swindle are shown in Figure 2, while the methods used by swindlers to appear as a legitimate seller to potential bidders are shown in the accompanying table.3
Buyers afraid of being swindled will stop participating, thus fewer buyers, meaning fewer sellers, and so on.
Our findings, although preliminary, could spell trouble for bidders in online auction sites, and ultimately the sites themselves. Even though the number of buyers and sellers has been rising rapidly over the last decade, any sudden shift in perception could reverse the cycle : Buyers afraid of being swindled will stop participating, thus fewer buyers, meaning fewer sellers, and so on.
We have several broad categories of recommendations based on our study. In this section, we review each category and give specific recommendations to reduce auction fraud along with the advantages and disadvantages of doing so. The broad categories are the following:
Note that the first category is probably the most important, and the first and last are the easiest to implement.
Information on sellers. Generally speaking, the more information is disclosed on sellers (and buyers) the easier it is for buyers and sellers to verify that they are dealing with a reputable entity, leading to a higher degree of confidence in the system. The problem now is that the system is based on self-reporting by both buyers and sellers. The auction house cannot guarantee that the information they provide is accurate.4 It is still the responsibility of the buyer and seller to verify the data.
Auction houses collect a significant amount of data. Making more of it public increases the chances that a transaction will be valid. Here is some data that should be posted for all reputation scores:
Advantages and disadvantages. The more information buyers and sellers have on each other, the better off they are. We should keep in mind that this additional information can be used by swindlers to better target their messages/offers to potential victims. Auction houses, on the other hand, have interests in protecting the identity of buyers and sellers, so it is a question of balance. All of these involve disclosing information about buyers and sellers (mainly sellers). In our opinion, however, legitimate sellers should be happy to reveal their information to benefit from the price premium of being reliable.5 This direct cost is negligible, involving hiring a few people to develop the algorithms and the software to display the new calculations.
Escrow. Extremely easy, seamless, and possibly mandatory escrow services should be available for all auctions. A small service charge could be added to the listing fee to facilitate the transaction. We propose the following modifications to the escrow systems currently in place:
Advantages and disadvantages. The advantages include a steep reduction in information asymmetries at the expense of swindlers. Information asymmetries exist when one party does not know everything relevant to a transaction that the other party knows. The schemes proposed here increase the knowledge of buyers about the quality (or even existence!) of the items to be sold. The disadvantages include greater transaction costs, even in the more mild versions and the possible exclusion of some buyers from the market (some buyers are willing to take a chance on being swindled for a lower pricethose buyers would be priced out of a market in which escrow was mandatory). One could imagine an extension of this in which only auctions closing above a certain dollar amount were forced to adopt some of these techniques.
The schemes proposed here increase the knowledge of buyers about the quality (or even existence!) of the items to be sold.
Performance Bond. To open an account as a seller or buyer, one needs a credit card. The auction house should charge the seller an amount equal to the amount of the sale (temporarily). Then, once the buyer gives the go-ahead, the auction house would reverse the charge.7 Auction sites already do this on the buyer side. The amount charged could also be based on the seller's reputation or other characteristics of the seller (such as, how many prior items sold at the current price range). This "seller escrow" would also benefit sellers without an established reputation. If the credit card transaction is rejected, than the auction house knows that this is not a legitimate seller and will remove his or her ID and items from the auction site. The auction site would also know immediately that they are probably dealing with a swindler.
The company would need to provide a cash flow management guarantee (return your money within, for example, one day), though, to protect those legitimate sellers who make a living selling on auctions. Even so, most sellers would only take a one-day hit because after the first day, they would be applying their performance bond to future auctions.
Advantages and disadvantages. The performance bond is a very strong level of protection and would sharply increase trust in the system as well as sharply reduce swindling. However, it is fundamentally a little unfair as it requires sellers to be relatively well-off, with enough credit on their credit card to pay for the items they already own. If someone was auctioning off items to raise money, say for a hospital bill, and they were already in debt, they would not be able to participate as a certified seller under this system. An alternative could be some kind of certification by the auction house for unusual circumstances, but this is likely to fail for two reasons: the auction houses would like to avoid certifying individual sellers for legal reasons; and the less "automated" the process, the more likely that actual swindlers would be able to pass off as legitimate customers. A third alternative could be the encouragement of third-party certifiers (or surety bond companies) such as BuySafe, which certify sellers for a commissionpaid by the seller on items soldand promise to compensate buyers if there is a problem.
Insurance. Appropriate insurance policies could encourage electronic commerce activities. Premia should be reasonable, and mechanisms designed to reduce or prevent collusion. The way it would work would be that the buyer would pay a premium to the insurance company and if there is a problem, the buyer makes a claim and is reimbursed. Here are a few possible recommendations for insurance:
Advantages and disadvantages. Smarter insurance policies would make the market fairer, especially if taking out a policy were mandatory. It would be fair because buyers would be hedged against absolute fraud and would not lose their money. In addition, the market for insuring fraud would become more efficient, because there are relatively few insurers and they would be scrutinizing the antecedents of fraud and pricing the risk accordingly as their own money would then be on the line if they are wrong. The disadvantage is that on the margin, the increased transaction cost would exclude some buyers. In addition, there could be a moral hazard problem in that buyers would be participating in riskier auctions (ones they estimate might have a fraudulent outcome) but they do not care as much because their insurance coverage provides a safety net.
Regulatory control. State or federal governments in the U.S. and elsewhere may want to make sure that the auction houses follow the rules set in their own descriptions of the mechanisms. There is a need for an agency that confirms they follow without bias the rules they advertise. Auction houses have at their disposal the ability to take advantage of situations in which they can make extra money at the expense of sellers and/or buyers, just as a real estate agent could conceivably buy and then resell a house rather than showing it to a potential buyer. We are not claiming that auction houses intervene in the auction markets frequently, but in fact who would know if they did or did not? Are the auction houses completely neutral? Some sort of government oversight might be helpful in this area.
In return for submitting to regulatory control (and for taking some proactive steps toward reducing fraud as outlined in this article), we propose that auction houses could be shielded from some forms of legal liability in fraud prevention. It could be that auction houses have resisted any action up until now for fear of legal reprisals, in other words, if they pursue better escrow or insurance policies, that might be considered an admission of guilt and open them up to lawsuits by unhappy auction winners.
Advantages and disadvantages. Without any a priori knowledge of biased interventions in their own auctions, we are not certain that instituting government oversight will create more problems than it solves. Still, adding government oversight will make the entire system more transparent. Further, if auction houses could be shielded from some legal liability through an oversight process, it might give them incentives to implement anti-fraud policies rather than simply claim they are a neutral market and are not at all responsible for members' trades.
Buyer precautions. There are many tools available on the Internet to help either party verify the information given by a seller. Examples:
Advantages and disadvantages. The advantages of buyer precautions are obvious, they reduce fraud and make it more difficult for crooks. They do not cost as much, if anything, to implement. The disadvantages are that not everyone would take advantage of them.
Auction houses today appear to be at a crossroads, with many people now losing confidence in the system :
"...do you have a reasonable expectation that the Mac G5 you are bidding on is going to be shipped to you? How about the Sony plasma 60-inch monitor? Or, the Sony PSP for $50 to $100 over retail? No, you don't. Now, you might say, 'Hey, moron! Why are you bidding on stuff like that on eBay? Don't you know that 95% of those auctions are fraudulent, especially the ones that only accept payment via Western Union?' Of course, I know. But what about all of the people who don't?"
Now is the time to act before the negative cycle mentioned at the beginning of this article develops in full swing. We hope that some of the recommendations listed here restore the public's confidence in the system. While it may be impossible to eliminate Internet auction fraud, at the very least we may be able to reduce it drastically and make it very expensive for those who persist.
2Although if we think that swindlers use fake U.S. addresses, we would not expect the location to be associated with swindling. Incidentally, broadband Internet phone servicewherein non-U.S. swindlers can obtain U.S. phone numbers and answer them anywhere in the worldwhile providing a very valuable service to legitimate customers, is actually exacerbating this problem.
3Buying and selling on the Internet is a very dynamic system that is evolving over time. For every fraud detection procedure that auction houses institute, swindlers adapt and deploy countermeasures that overcome the new defenses. We realize that we might not have covered every method; if a reader knows of anything else, please send it to us so we can continue with our larger follow-up study.
4Using tools available on the Net it is possible to verify a significant part of the information needed to establish the person or company identity, for example, does a person with that name exist in his address? Is the phone number associated with him? Is such a company declared and filed with his local authorities? What is his educational background? Does he have a family? With whom did he deal in the past? In some cases, what is his credit score? The amount of useful data that can be collected on the Internet pertaining to an individual is surprising.
5Revealing the sellers identity on the auction site could be an option that the seller selects, thus giving the indication that he is legitimate. Sellers of very expensive items (cars), list in the body of the listing information that allows potential buyers to verify that they do exist and do business (their dealership title, address, phone number, vehicle location and stock number, VIN, and many pictures of the car).
©2008 ACM 0001-0782/08/0500 $5.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2008 ACM, Inc.