Other Approaches

The explicit assessment of value functions, which the scoring function approach requires, has been criticized by several researchers over the years (e.g. Simon 1955, Larichev, 1984, and Korhonen and Wallenius, 1996). It requires practice and expertise to be able to express one’s preferences in the form of a value function. In the auction setting, the assessment of the auction owner’s preferences should be easy in order to entice managers to resort to auctions in the procurement process. The applicability of the scoring function method is thus questionable. Bichler (2000) added a decision-aiding tool to help auction owners construct their value functions. However, he does not describe the tool. If the process is not transparent and understandable to the user, it will not evoke trust in the procurement manager and she might decide not to use the auction system.

The focus of recent studies in multi-attribute auctions has diverted from the focus of traditional auction research. Now the focus is not as much the efficiency or the revenue (utility) equivalency of auction mechanisms, but the functionality of the designed auction. Functionality refers here to the ease of use for both the seller and the buyers. Even though the primary goal is not to design an auction that maximizes the auctioneer’s revenue (utility), it is, of course, important that the auctioneer receives a satisfactory utility. Multi-attribute auctions can be seen as cooperative negotiations (see e.g. Guttman and Maes, 1998), as there can be opportunities for joint gains. In a way, multi-attribute auctions resemble traditional one-on-one negotiations. Therefore, some multi-attribute auction design builds upon negotiation theory, e.g. the The Leap Frog Method and the Auction Owner Controlled Bid Mechanism suggested by Teich, Wallenius and Wallenius (1999).

Cripps and Ireland (1994) propose a method, which uses quality thresholds. This would eliminate all the other attributes besides price, and render the auction to the price-only situation. They consider three specific designs. In (a) a price-only auction is

held only after bidders have submitted their quality plans (which contain information on all non-price attributes) and the plans have been accepted. In (b) the auction is held first, and quality plans are requested in the order indicated by the auction results, starting with the winner. The first bidder, whose quality plan is approved, obtains the contract. In (c) price and quality plans are submitted simultaneously, and the contract is awarded to the best priced (i.e. the cheapest deal in the reverse auction setting) plan that satisfies the predetermined quality requirements.

Teich, Wallenius, Wallenius and Zaitsev (2001 and 2006) have implemented the

“pricing out” method in their Internet-based hybrid auction called NegotiAuction^TM. Pricing out can be used without having to explicitly formulate the auctioneer’s value function. Instead, it probes into the implicit preferences of decision makers. This makes it a popular approach among decision analysts. It is, however, possible to construct a value function with pricing out, but Teich et al. (2001, 2006) choose not to do so.

Pricing out can be used in all situations, where there is a natural monetary attribute (price or cost) attached to the auctioned asset. Borcherding, Eppel, and von Winterfeldt (1991) compared pricing out with three other value elicitation techniques.

They concluded that weights generated with pricing out corresponded closest with weights generated externally by a group of experts.

Keeney and Raiffa (1976) provide a general description of the pricing out method. The underlying idea is to express the decision maker’s (in this case the auction owner’s) preferences over multiple attributes in monetary terms. Assume that there is a monetary attribute M and n different non-monetary attributes X₁, X_2, … , X_n related to a product. Lower case letters m and x₁, … , x_n denote the values the attributes can assume. In pricing out the auction owner is asked to identify the monetary value m^* for a bundle x^*=(x₁^*, … , x_n^*) which makes her indifferent between the bundle (m^*, x^*) and a predetermined reference bundle (m₀, x₀). That is, (m^*, x^*) ~ (m⁰, x⁰). The difference m^* - m0 depicts the auctioneer’s willingness to pay for the possibility of transforming the bundle x⁰ into x^*.

The approach explained above becomes tedious and time-consuming when the number of vectors that need to be evaluated is large. Then, it pays off to make simplifying assumptions. Keeney and Raiffa (1976) state that pricing out can be made easier when

1) The difference between m^* and m⁰ (i.e. the willingness to pay) does not functionally depend on the base value of m⁰

2) The monetary attribute M and attribute X_i as a pair are preferentially independent of the complementary set of attributes

When these assumptions apply, the pricing out can be done individually for each attribute. The method is the same as in the general case described above, but here the value of only one attribute will be changed at a time. For each attribute X_i the auctioneer is asked to state a monetary value m^* that will satisfy the indifference equation

(m^*, x₁⁰, … , x_i-1⁰, x_i^*, x_i+1⁰, … , x_n⁰) ~ (m⁰, x⁰) (4) Thus the auctioneer only has to make n such assessments instead of pricing out all possible combinations of the attributes X. This simplifies the preference elicitation process and formulates it in such a way that it is easy for the auctioneer to make the assessments.

Pricing out also provides computational advantages, because it reduces the bids to two-dimensional vectors containing only price and quantity components. The bidders still submit multi-dimensional bids, but all other attributes are “priced out” before the winner determination problem is solved.

The various preference elicitation methods described above try to achieve two goals:

the realistic and truthful description of the preferences of buyers and sellers and the ease of use of the method for all participants. Unfortunately, most of the time these goals conflict. The more elaborate the preference elicitation scheme is, the more difficult it is for a novice to use. The scoring method clearly emphasizes the first goal and is therefore more suitable for theoretical studies. The rest of the above mentioned approaches prioritize the usability of the methods attempting to generate mechanisms that could be implemented in practice. Regardless of the method in question, it is clear

that in the multiple-issue case, bidding becomes more difficult for the bidders, especially inexperienced ones. Also, it is more difficult for the auction owner to set up an auction that would produce the kind of results that would match her true preferences. Hence, all sorts of decision aid tools become important in auctions (Bichler, 2000), and decision making theory becomes relevant for auction theory. The introduction of Internet auctions provides an excellent medium to include decision support with auctions, as will be discussed later in Chapter 5.