-
Chapter 5: Trust and Reputation in Internet Auctions
- Russell Sage Foundation
- Chapter
- Additional Information
Chapter 5 Trust and Reputation in Internet Auctions Andreas Diekmann, Ben Jann, and David Wyder E conomic exchange between anonymous actors is risky for all interacting parties. Whether in barter or in sale against cash, in a bilateral exchange situation both actors have to choose between being more or less cooperative or acting fraudulently. A seller, for example , needs to decide whether to deliver at all, to deliver good quality, or to deliver bad quality, and a buyer may choose to evade, reduce, or delay the payment. It is well known that such cooperation problems can be solved by repeated interactions if the shadow of the future, that is, the expectation and valuation of future transactions, is sufficiently large (Axelrod 1984; for a survey, see Diekmann und Lindenberg 2001). However , no such temporal embeddedness occurs in single transactions (Raub and Weesie 2000). Hence it is likely that both actors behave uncooperatively . Internet auctions, characterized by anonymity and nonrepetition of transaction, closely correspond to this type of interaction. Sellers and buyers may adopt virtual identities, that is, they may act under fictitious names and fake addresses, and it is evident that to realize mutually satisfactory exchanges a basic trust problem must be overcome.1 In terms of game theory, an Internet auction with simultaneous transaction corresponds to the ideal type of a single one-shot prisoner’s dilemma. If, however, the actors fulfill their obligations sequentially such that the second actor can condition his move on the action of the first, a sequential prisoner’s dilemma or trust game is played. A single exchange between anonymous actors is a precarious situation and gives reason for the prediction that both parties will strongly tend toward fraudulent behavior without intervention by a central au139 thority. Therefore, one would expect cheating in Internet transactions and unstable markets that collapse rapidly or fail to emerge despite demand . Contrary to expectation, several Internet auction platforms such as eBay, QXL ricardo (now Tradus), or Amazon have been successful for years. Apparently, these markets do not erode due to lack of mutual trust. Furthermore, cheating in Internet auctions seems to be relatively rare. Peter Kollock mentioned early figures by eBay according to which only twenty-seven cases of fraud have been reported out of 2 million auctions between May and August 1997 (1999). The National Fraud Information Center/Internet Fraud Watch (NFIC/IFW), a project of the National Consumers League of the United States, is concerned with registering cases of Internet fraud and forwarding them to the appropriate law enforcement agencies. According to the Internet Fraud Statistics of the NFIC/IFW, the majority of all registered cases of Internet fraud around the time of our study occurred in Internet auctions.2 The average monetary loss per Internet auction fraud victim amounted to between $300 and $400. Even if the NFIC/IFW statistics underestimate the actual crime rate, risk of fraud is relatively low given the millions of transactions handled by auction platforms such as eBay and Tradus.3 The reason for the success of these Internet markets is a simple institutional rule. Both actors participating in a transaction, buyer and seller, are advised to rate each other after the deal has been completed. That is, the actors may valuate the other party’s business conduct by assigning marks and verbal statements, and these assessments are open to anyone who is interested. Thus a potential buyer can browse a seller’s list of received ratings from previous transactions before placing a bid. To simplify matters, auction platforms usually also provide summary reputation indices based on the single ratings. In the time we collected our data, ricardo.ch declared the average number of stars and the number of transactions on which this average measure was based. Additionally, separate statistics for positive (four or five stars), neutral (three stars), and negative (one or two stars) assessments were provided for the most recent transactions (figure 5.1). Although the rating process is reciprocal, that is, seller and buyer can both submit a rating in a given transaction, the assessments given to sellers seem to be more important, because bidders may pick sellers by their reputation, but sellers may not choose buyers. Trust in exchange situations arises from learning from past behavior of the contracting partner and from control, that is, the possibility to impose sanctions in the case of uncooperative behavior (Buskens and Raub 2004). From the viewpoint of the buyer, both elements, learning and control, are inherent components of...