بهره وری عملیاتی مکانیسم های غیرمتمرکز مزایده اینترنتی

The recent consumer-to-consumer (C2C) Internet auction boom at eBay, Yahoo, Amazon, and other sites has added new theoretical challenges to the science of auctions. This paper uses experiments with economically-motivated human subjects to measure the operational efficiency of decentralized Internet auction mechanisms as they compare to centralized double auction mechanisms. Subjects are recruited randomly from the undergraduate population of a large urban university to be buyers or sellers in a simulated trading environment. They are randomly assigned costs and values for 10 units of a virtual product. During the experiment subjects trade these units through computer terminals in auctions similar to those held on eBay and generate profits, which subjects receive at the end of the experiment. The paper uses data from this experiment and previous laboratory experiments to compare operational efficiency and convergence pattern of prices to equilibrium levels in continuous double auctions versus online Internet auctions based on two variables: auction size and time. Experimental results suggest that, because of their decentralized nature, Internet auctions require a few more participants and more time to achieve operational efficiency levels than centralized markets which use continuous double auction mechanisms.

Over the last decade online consumer-to-consumer (C2C) auctions have had an enormous impact on business the world over. Yet we still know very little about how efficiently goods are being exchanged online and whether and how the efficiency of the online trading process could be improved. This is in parallel with the relative lack of studies that reliably measure efficiency even in traditional offline auctions. At the same time efficiency is often one of the major criteria when an agency is choosing between possible alternative mechanisms for selling various property rights or other products (e.g. see Cramton, 1998 and Cox et al., 2002). Tracking the efficiency of an auction mechanism has also important practical significance for online auction managers (Gallien and Gupta, 2007, Kauffman et al., in press and Caldentey and Vulcano, 2007). On the web the major source of revenue for auction marketplaces is the commissions and listing fees. Commissions are usually a percentage of the transaction price. In order to maximize revenue, online auction managers should maximize transaction prices and transaction volume. However, if the process of raising prices also results in much lower buyer surplus, then many buyers would be turned away from the auction website to other alternatives. In order to improve profitability without hurting buyers, online auction managers can implement policies that increase transaction prices and auction efficiency at the same time and at least at approximately the same rate. This would guarantee that buyers are not hurt in the process of increasing auction website revenue. This is why auction managers should be concerned about efficiency when they make changes to auction rules that might influence auction prices and performance (Wenyan and Bolivar 2008). There are two different definitions of market efficiency that have been used to assess how well auctions in general and Internet auctions in particular perform. The first type of market efficiency is known as operational (or allocative) efficiency. This efficiency is defined as a percentage of the maximum possible surplus extracted by a market institution while demand and supply are being matched (see Parsons et al. 2006). This idea of efficiency works well for final products – or products that have well-defined production costs that the sellers incur and also have some intrinsic usually heterogeneous values to the buyers (see Milgrom and Weber 1982; also Krishna 2002, and Klemperer 2004). Both buyers and sellers need to perform only one transaction in order to enjoy gains from trade. The difference between the transaction price and the production and other costs is the seller surplus, and the difference between the buyer’s value and the transaction price is the buyer surplus. The sum of these two surpluses is the total surplus, and operational efficiency is the ratio between the total realized surplus and the total possible surplus. This idea of efficiency allows establishing efficiency baselines, ranking different auction mechanisms, and makes possible the estimation of the effect of a change in a certain market variable that is under a market designer’s control. This type of efficiency is hard to estimate using market data except in some limited circumstances (see Kang and Puller, 2008, Hortacsu, 2002 and Gopal et al., 2007) because much of the information about real costs and values is never fully revealed by market participants. An easy way to see how operational efficiency is impacted by market variables is through laboratory experiments with economically-motivated human subjects ( Smith, 2002 and Smith, 2003) in which values and costs are directly induced by the experimenter. The second idea of efficiency, which we call informational efficiency, exists when there is no potential for arbitrage in the market. This idea of efficiency is usually used to assess how well financial markets perform (see Fama 1991). The idea is useful for financial markets because financial paper does not have intrinsic value, that is, in order to realize profits, one has to buy a financial product and then sell it later. Thus the value of a financial product depends on the future expectations of all market participants. Market participants’ values for products similar to financial paper are common or correlated. Informational efficiency is thus quite relevant in financial and other markets where market participants are expected to re-trade an item before they can realize a profit. Financial economists have developed methods to detect if a market is informationally efficient by using past price data (see Davis 2008 for a review). There have already been several studies that have used this methodology to find that current C2C auctions are not informationally efficient (see Kauffman et al., in press). What are the factors that impact market efficiency? Classical economic theory suggests that one of the most important variables that can affect efficiency is the number of participants in a market mechanism (see MacKenzie et al. 2007). This is also termed market size. More recently modern auction theory has devoted much attention also to the importance of the market mechanism which is being used to match supply and demand (see Krishna, 2002, Klemperer, 2004 and Smith, 2003). We know the structure and rules of currently available Internet auction mechanisms. However, we do not know exactly how inefficient the auctions are and how their efficiency is impacted by an increase in the number of auction participants. This paper is an initial attempt to fill this gap in current research. In the new context of Internet auction mechanisms, this study asks the following research questions: • How is the operational efficiency of online auctions influenced by auction size? • How is the relationship between auction size and efficiency different in decentralized Internet auction mechanisms versus more centralized market mechanisms like continuous double auctions? •How do prices converge over time to competitive levels in more decentralized mechanisms like Internet auction mechanisms versus centralized market mechanisms like continuous double auctions? To address these questions the paper uses an exploratory laboratory study with economically-motivated human subjects. The experiment presented here was conducted in the summer of 2001 and is among the first laboratory experiments involving human subjects that tries to simulate the economic environment surrounding online auction mechanisms for final goods. The main experimental finding reported here is that Internet auctions need more than seven buyer visits per auction in order for auction prices to reach competitive levels. It also turns out that Internet auction mechanisms require more time than centralized markets to achieve high efficiency levels. This paper makes several contributions to theory and practice. First, it shows that a basic principle from economic theory about the relationship between market size and market efficiency applies to Internet auctions. Second, it extends previous experimental work in auctions to show that Internet auction mechanisms are different in their convergence properties from centralized double-auction mechanisms. Third, the paper can serve as a guideline to online auction designers as to how operational efficiency could be measured in the laboratory and describes a method to establish an efficiency baseline, change certain auction variables and estimate the effect of that change on the operational efficiency of the auction mechanism being tested. Lastly, the paper uses the reported research findings to suggest ways in which operational efficiency of online auctions could be improved. The article is organized in the following way: Section 2 provides an overview of previous research related to C2C Internet auctions and describes in detail the main differences between the centralized commodity markets experimentally tested by Smith and some of the most popular current C2C Internet auctions like eBay. Section 3 discusses the methodology and Section 4 describes the main features of the experimental design. Section 5 reports the experimental results, Section 6 discusses their implications. Section 7 provides a summary of the limitations and the conclusion.

This paper compares the operational efficiency of Internet consumer-to-consumer (C2C) auctions to the operational efficiency of the continuous double auction (CDA), a market institution whose efficiency properties have been studied extensively in the laboratory. The author chose the experimental method to explore this comparison because this is the only method which allows direct observation of operational efficiencies and exact deviation from equilibrium prices. The results of the study demonstrate because of four basic features of C2C auctions that continuous double auctions do not possess, C2C auctions achieve lower market operational efficiency levels than the continuous double auction under similar economic environments. These four features are: the distributed localization of demand and supply, the incompleteness of market information, the increased importance of time-related cost, and the asymmetric strategic relationship between buyers and sellers. The experiment reported here also explores the relationship between operational efficiency, market size and time in Internet auctions. The experimental results are compared to naturally occurring auction data to establish similarity and then to earlier experiments that look at similar relationships in continuous double auctions with open outcry of bids and offers. The experimental results confirm that Smith’s findings pertaining to centralized institutions of exchange do not apply with the same strength to decentralized C2C auctions on the Internet. Generally prices converge much more slowly to the equilibrium level in Internet auctions. There is some downward stickiness in prices that might be erased with the presence of more bidders. It appears that an Internet auction needs to be visited by at least seven bidders and needs to have at least four bids from different bidders in order for the prices to reach competitive levels and for operational efficiency to move closer to 100%. From an auction design perspective the experimental results reaffirm the claim that laboratory experiments are a cheap and reliable method to duplicate online auction behavior. The study shows that one can reliably trace causal relationships between market variables by manipulating experimental parameters, establishing efficiency baselines, and testing a wide range of design options before an auction is implemented. The results also help us suggest changes in online auction design that have the potential to improve the operational efficiency of online auctions. The experimental design outlined above does not simulate all practical aspects of Internet auction mechanisms. For example, all auctions in the experiment happen within the span of an hour while auctions on eBay take at least 3 days unless the buy-it-now option is exercised. Roth and Ockenfels (2002) suggest that this limitation might not be that restrictive because the relationship between bid timing and auction length is fractal and thus the graph of bid timing has a similar shape irrespective of the auction length. The same relationship might be true for the bid-prices in C2C Internet auctions and future research in this direction will throw more light on this yet unexplored property of the time/bid-price relationship. The experiment also could not probe Internet auction environments with more potential visits and bids per auction in order to discover the auction size required to erase the collusive pressures on the price. Other necessary additions to this experimental design that we omitted include flexibility of auction rules, reputation, shill bidding, and agent usage. Testing the effect of these additional features on prices and efficiency should be performed to get a complete picture. An interesting exercise would be to investigate the relationship between informational and operational efficiency in environments that represent a mix of financial markets and markets for final products. Another possible concern is the level of experience of subjects in the laboratory as compared to bidders in Internet auctions. Obviously participants in online auctions are a very diverse group of people and might act differently than undergraduate subjects. This study, however, is comparative in nature. Efficiencies of two different mechanisms are measured while the experience of the experimental subjects is the same. Subjects in both the experiment reported here and these reported in Smith, 1962 and Smith et al., 1982 were undergraduate students from a large public university in the US. Another limitation of this laboratory study was the size of the laboratory. As explained earlier, only 26 computers were available for experimental subjects. Because of this limitation alternative ways to measure auction size were not viable. Clearly a better picture of the relationship between the several important variables investigated here could be obtained by a larger study. In addition the study does not model the winner’s curse – a phenomenon that has been detected in online auctions. Subjects’ values and costs in this experiment are independent and private. Winner’s curse happens when market participant’s values are affiliated or common. Since independent private value environments are simpler from a market participant’s and an experimenter’s viewpoint than affiliated or common value environments, they are a good starting point for this experimental study. Later experiments could model affiliated or common values in addition to private values. Given the results of our study, a plausible expectation is that online auction mechanisms would perform worse than continuous double auction mechanisms in more complex environments as well but this is something that is still to be tested in the laboratory.