|
Article Excerpt
1. Introduction
Ariba, (1) a B2B market-maker, convenes electronic reverse auctions (i.e., market sessions) at the request of buyers. Among other attributes, buyers determine the market-transparency scheme, or information-revelation policy, to be used in the market session. A revelation policy determines the nature of information about bids--winning bids, number of bidders, etc.--that is revealed to geographically dispersed sellers in a market, at the beginning, during, or at the end of a market-session. At one end of the spectrum of available policies, the buyer can choose to accept sealed bids and inform each seller only about whether it won or lost in a given market session. Under this policy, competitive information is not revealed to sellers. At the other end, the buyer can choose a revelation policy that allows sellers to observe the bids submitted by their opponents in real time and react. Under this policy, sellers are aware of the number of other sellers, their bidding patterns, and winning bid prices. The implications of such revelation policies are significant in marketplaces wherein the same set of sellers repeatedly compete over multiple market sessions. Because the information content is different across revelation policies, the policy chosen affects what sellers learn and how they bid in the future, which in turn determines buyer surplus. In this paper, we analyze the impact of such information-revelation policies on buyer surplus.
Prior work, such as Wilson (1977), Winkler and Brooks (1980), Grossman and Stiglitz (1980), and Milgrom and Weber (1982) has considered the information impact on market performance. However, this paper focuses on the sellers' attempts to learn about the number of competitors they face, and the consequences for buyers. Such a type of uncertainty, referred to as the market-structure uncertainty, arises naturally in information-technology-enabled marketplaces and poses decision problems for managers. In our model, the buyer chooses a reservation price and the information-revelation policy for a market session, while sellers have to determine optimal bidding strategies. These decision problems need to take into account the behavioral response of other parties in the market under incomplete information. It is not ex ante clear if more information regarding market-structure uncertainty can make a player worse off or not (Gibbons 1992). In some cases, this can increase the intensity of competition. We focus on a related but different issue: How the need to acquire information affects behavior, and in turn, how such behavior affects market outcomes. To the best of our knowledge, the problem addressed in this paper has not been studied in prior literature. By addressing this problem, our work contributes to the growing body of information-systems (IS) literature on design issues related to e-markets (e.g., Bakos and Nault 1997, Zhu 2004).
In this paper, we consider the following revelation policies:
1. Complete market-structure information setting (CIS): All quotes are revealed to all participating sellers, implying that all sellers learn about the structure of the market. This is similar to municipal construction contracting (Thomas 1996).
2. Incomplete market-structure information setting (IIS): The only information revealed to participating sellers at the end of each market session is the winner's quote. This means that the winning seller, who is already aware of his or her own bid, does not learn anything about the market structure. However, losing bidders, if they exist, learn about the presence of at least one competitor and can bid accordingly in future market sessions. This setting is similar to government procurement auctions, where statutes require the winner's bid to be revealed at the end of the market-session (McAfee and McMillan 1988).
Note that these policies are but two of the many policies available to a buyer in a real-world marketplace such as those hosted by Ariba. Further, these specific policies are commonly adopted in traditional marketplaces (Thomas 1996).
We demonstrate that buyers benefit from choosing CIS over IIS. In other words, the policy that generates the least amount of market-structure uncertainty for the sellers always maximizes buyer surplus. We further analyze to show how bidders' reactions to overcome uncertainty differ with the nature of uncertainty, and how those reactions impact buyer surplus. Finally, although our work is motivated by a real-world electronic marketplace hosted by Ariba, the results discussed in this paper are applicable to any reverse market setting that can create these different information regimes.
The paper is organized in the following manner. Section 2 provides the literature review and [section]3 describes the problem context and models it. In [section]4, we solve for the equilibrium and compare CIS and IIS. Section 5 shows that the basic results are robust to relaxing a number of simplifying assumptions. The implications of the results and the contribution of this paper are summarized in [section]6. Finally in [section]7, we conclude.
2. Literature Review
To our knowledge, revelation policies in single-sided auctions have been studied only in Thomas (1996), Koppius (2002), and Zhu (2004). Thomas (1996) compares CIS and IIS in a two-seller setting where each seller is certain about the presence of its opponent but is uncertain about its opponent's cost structure. Using this model, he demonstrates that the setting equivalent to our CIS generates higher buyer surplus than IIS. (2) Our paper is different from Thomas (1996) in that sellers in our model are certain about their opponent's cost structure but are uncertain about the presence of their opponents.
The paper by Koppius (2002) uses experimental data to compare the impact of revelation policies on sellers' profits in a multidimensional auction. They show that the policy with the least level of uncertainty for the sellers about the weights that the buyer places on each attribute, generates the highest profit for the bidders. In his paper, Koppius (2002) does not take into account the uncertainty about the number of competitors, which is the focus of our paper.
Zhu (2004) analyzes the informational effects of exchanges but focuses on participants' inferences about their rivals' costs. He studies the incentives of the firms to join the exchange under both price and quantity competitions by considering a model where firms participating in the exchange can observe the costs of other participants. In this model, only a few firms participate because the loss that high-cost firms suffer when their costs are revealed to others outweighs the benefit of learning about the costs of others. Our paper is different from Zhu (2004) in many respects. First, our paper focuses on market-structure uncertainty whereas Zhu (2004) focuses on uncertainty about opponent's cost. Second, we focus on learning across market sessions, which is not the focus in Zhu (2004).
There are many papers that address learning issues in auctions. Most prior research work in this domain can be categorized along two dimensions. The first dimension focuses on the bidders' decision problem of whether or not to invest to learn the value of the auctioned item so as to be able to use that information to participate in the auction. One of the earliest papers to explore this idea is Schweizer and Ungern-Sternberg (1983), which models bidders having random valuations for the product. The bidders can narrow the intervals of this valuation by incurring a learning cost. Schweizer and Ungern-Sternberg (1983) study the behavior of bidders using simulations. Milgrom (1981) studies a two-stage game, in which a bidder can learn the common value of the auctioned item by incurring a cost. Milgrom (1981) characterizes the rational expectation equilibrium for sellers in this game. These models are tested experimentally by Guzman and Kolstad (1997), who also analyze the effect of variations in parameters such as information costs and levels of valuation uncertainty on bidding behavior.
The second dimension of research on learning issues in auctions focuses on bidding behavior in multiperiod games (auctions) where bidders who are certain about the presence of their opponents, are uncertain about their opponents' attributes, including their payoffs. Thomas (1996) (discussed earlier in this section), Snir et al. (1998), and Dekel et al. (2004) study equilibrium strategies in this context. Snir et al. (1998) demonstrate that in a repeated-game setting where the payoffs of the opponent are not known, the equilibrium eventually converges to that of a single period game with known payoffs. Dekel et al. (2004) analyze Bayesian games where payoffs of the opponent are unknown and discuss the restrictive assumptions needed about priors to justify the Nash equilibrium. Although we also study learning across auctions, our focus is on the impact of information-revelation policies on learning.
Our work addresses the impact of revelation policies under market-structure uncertainty. Some prior work has modelled market-structure uncertainty as endogenous participation, e.g., Harstad (1990) and Janssen and Rasmusen (2002). Harstad (1990) considers a common-value auction framework and observes that buyers prefer few rather than many bidders to participate. Janssen and Rasmusen (2002) consider a market with endogenous participation to compare the Cournot and Bertrand pricing games as sellers face uncertainty about the number of competitors. Both these papers restrict their focus to single-period auction games where learning from one auction to the other is irrelevant. We also extend our model to allow for endogenous participation but our focus is how bidders alter their first-period bids to learn about the market structure. Thus, one contribution of our model is to extend Janssen and Rasmusen (2002) to consider CIS and IIS. The signal-jamming model, analyzed by Fudenberg and Tirole (1986), is the closest to ours. In that model, one informed player attempts to take advantage of its rival's lack of information about market and cost conditions. The lack of market information leads to lower prices in the first period as the (informed) firm attempts to prevent its rival from learning whether the market is profitable or not. By contrast, in our model, all firms are ex ante identical, and equally (un)informed about the possible rivals.
2.1. Finance Literature Review
Revelation policies in financial markets are referred to as trade transparencies in the market microstructure literature. Depending on whether information about outstanding orders or sscompleted orders are revealed, these...
|
