Using experimental auctions for marketing applications: A discussion

Journal of Agricultural and Applied Economics, Aug 2003 by Lusk, Jayson L

In CA, subjects are shown several different product scenarios or profiles, where the attributes of a good (such as price, packaging, brand, etc.) are varied across scenario. Subjects are asked to rank or rate the scenarios or are asked to choose which scenario or product profile is most desired. For the remainder of the present article, the term CA will be used to refer to the method of choosing the one product that is most preferred.3 Most CA applications are done in a hypothetical context and, as such, can be viewed as a type of CV; however, CA can be readily used in a non-hypothetical context to mitigate problems with hypothetical bias (e.g., Lusk and Schroeder). CA has been widely used in the marketing literature, and its advantages over EA include (a) CA elicits responses in a manner that closely mimics actual shopping behavior, whereas EA require subjects to formulate bids in a manner that is unfamiliar to most subjects; and (b) with CA it is straightforward to estimate substitutability between multiple goods and attributes, which can be difficult or impossible with EA. Advantages of EA over CA include (a) EA elicit willingness-to-pay values for each individual, whereas willingness to pay must be indirectly inferred in CA from estimated utility functions with particular functional forms; (b) some studies have suggested that individuals' behavior in CA can be inconsistent or contingent on design parameters (e.g., DeShazo and Fermo; Swait and Adamowicz); and (c) it is relatively easy to model determinants of willingness to pay from EA, but this task can be quite difficult with CA.

A few of studies have compared willingness-to-pay values from CV, CA, and EA. Balistreri et al. found that bids from an English auction were significantly lower than those from both a hypothetical open-ended CV question and a hypothetical dichotomous choice CV question, which supports the extant literature on hypothetical bias. Frykblom and Shogren compared responses to a nonhypothetical dichotomous choice question to bids from a Vickrey second-price auction. They could not reject the hypothesis that willingness to pay elicited via the dichotomous choice question was equal to willingness to pay from the second-price auction. Lusk and Schroeder compared results from several EA with those from a CA. They found that bids from the EA were significantly lower than those implied from the CA. They also found that that the EA predicted much lower market shares for purchased goods that those observed in the CA.

In sum, EA are simply a tool that can be used to assist firms with product adoption and pricing decisions. EA are especially useful in situations in which one is interested in modeling determinants of valuations. EA are also noteworthy because of their reliance on an active market environment that is absent from most other marketing research techniques. Whether EA should be used for a particular application depends on the particular research objectives, and no doubt EA are not the best method to use for every problem. For example, if one needs to estimate valuation estimates that can be generalized to a national sample, conducting enough EA to generate national representativeness may be infeasible relative to a hypothetical CV survey. That said, EA have some distinct advantages over other marketing research methods, and agricultural economists interested in nonmarket valuation would be well served to become knowledgeable about EA.