reduce auction fraud along with the advantages and disadvantages of doing so. The broad categories are the following:

 

1. Increase the information shared by auction houses on buyers and sellers.

2. Make the use of legitimate escrow services extremely easy, and possibly mandatory.

3. Charge the seller an amount equal to the amount of the sale (temporarily).

4. Develop sensible insurance policies.

5. Institute regulatory control over the auction houses. _Shill

6. Encourage buyers to ^bidding protect themselves.

Auction “winner”

Note that the first cate- Initial Initial shipment non-

gory is probably thetakes delivery most important, and ^{places of item} the first and last are the easiest to implement.

Information on sellers. Generally speaking, the more information is Empty disclosed on sellers (and ^{box /} _worthless 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:

Fake escrow company

Product with inferior component sent

Damaged / nonfunctional product sent

Wait for Y (~ 14) more days

Promised product delivered but value has declined

Alternate product with inferior component

Figure 2. Possible Internet auction swindling methods.

• The percentage of positive responses of the total

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.

number of transactions for sellers (some sites began doing this in 2003) rather than the absolute number. In many cases, the auction houses only reveal the number of positive scores, thus if someone sells 1,000 items, of which 100 are positive, she or he would have a higher score than someone with 99 positives out of 99 auctions.

• As mentioned in the table and discussed in Chua
and Wareham [ 2], some sellers sell (or buy) a
large number of cheaper items to establish their
reputation. To address
this, we propose posting
_Auction the average selling price
“^loser” (or a distribution of
selling prices) of all the
Unsolicited seller’s previous auc-
^proposal _{for same} tions. It would be even
^item better to divide it into
months and also by
^{Fake Fake US} item category.
shipping business
company address • Separate statistics for
selling vs. purchasing.
This reduces the
Refund No chances that a seller can
^{offered refund,} _{after X seller} establish a positive rep-
^{(~ 45) days keeps} _cash utation by simply buy-
ing many low-cost
items, which he subse-
^Accept quently sells thus dou-
refund
immediately, bling his positive score.
paid in 60
days • Seller activity history
(not score) as a func-
^Refund _requested, tion of time. The pat-
^{paid in 60} _days tern of activity and
dollar value can provide
clues to participants.

• The number of distinct users (buyers) that participated in creating a reputation history for a seller.

• Statistics on the number and percentage of cases in which the seller was not paid, or buyer did not pay (indicates a shill bidder running up the price if it happens too often).

• Information on the seller, for example, listing the physical (as opposed to electronic) address for payment at the end of the auction. Such physical information can help determine if this is a legitimate seller or a swindler. In addition, there are Internet-based tools that could be used to disin-guish real from purely virtual sellers. Auction houses could accomplish this by sending a physical letter to the postal address of the registered user, much as a bank sends a PIN under separate cover to a postal address.