Using ex ante approaches to obtain credible signals for value in contingent markets: evidence from the field.

Our field experiment was conducted on the floor of a sports memorabilia show in Tucson, Arizona. As discussed in previous work (e.g., List 2001), with the rise in popularity of sports cards and memorabilia in the past two decades, markets have naturally arisen that allow for the interaction of buyers and sellers. The physical marketplace is typically a gymnasium or hotel conference center. When the market opens, consumers mill around the marketplace bargaining with dealers, who have their merchandise prominently displayed. The duration of a typical sports card show is a weekend, and a lucrative show may provide any given dealer hundreds of exchange opportunities (buying, selling, and trading of goods).

On the weekend in which we ran our field experiment, we approached attendees as they entered the sports card show and inquired about their interest in participating in an experiment. The interceptor explained to each potential subject that they would receive $10 for showing up if they decided to participate. Upon obtaining an agreement to participate, the interceptor informed the subject of the time and place of the experiment (a reserved room in the hotel conference center). Each subject was allocated to one, and only one, of the eight sessions. (As described below, each of the eight sessions represented a distinct treatment.)

Upon arrival to the experimental session, individuals signed a consent form upon which they agreed to abide by the rules of the experiment, received their $10 show-up payment, and were given experiment instructions. Depending on the session in which they participated, they were allocated randomly to one of the eight treatments summarized in table 1. Table 1 presents a summary of our 2 x 4 experimental design and provides sample sizes in each treatment. Table 1 can be read as follows: columns indicate treatment type--hypothetical, hypothetical-with-cheaptalk, consequential, or real; rows indicate pricing sequence--$5/$10 (pricing sequence A), or $10/$5 (pricing sequence B).

Taking one of these treatments as an example, consider an excerpt from the instructions of the real treatment (full instructions are available upon request):

Welcome to Lister's Referendum. Today you

have the opportunity to vote on whether 'Mr.

Twister,' this small metal box, will be 'funded.'

If 'Mr. Twister' is funded, I will turn the handle

and n (the amount of people in the room)

ticket stubs dated October 12, 1997, which

were issued for the game in which Barry

Sanders passed Jim Brown for the number

2 spot in the NFL all-time rushing yardage,

will be distributed--one to each participant

(illustrate). To fund 'Mr. Twister,' all of you

will have to pay $X.

We utilized a referendum vote for the provision of a public good as our value elicitation mechanism, as this is a common method utilized in field applications of stated preference methods and closed-ended mechanisms were recommended by the NOAA panel on contingent valuation (Arrow et al. 1993). As implied above, we obtain voting responses from each subject for each of two price levels (i.e., $X was $5 in one question and $10 in the other). For example, subjects in the real treatment pricing sequence A first provide a response for the $5 question and then for the $10 question, regardless of how the group responded to the initial $5 offer. (6) By varying the price level in such a manner, our data allow for within and between comparisons of the effect of offer price on voting behavior, in addition to between comparisons of treatment effects. We change the ordering of the offer prices to test for sequencing effects, as exhibited in the rows of table 1.

Our referendum mechanism operates on a simple majority vote for n identical pieces of sports memorabilia (where n equals the number of subjects). In the real treatment, these n private goods are provided to everyone in the experiment if a majority of the subjects vote to "fund" the public good "Mr. Twister," while provision is made in the consequential treatment if the referendum vote is binding (determined by random outcome from known probability distribution). In this way, the n private goods simulate the provision of a public good--no one is excluded from the provision of the sports memorabilia (or lack thereof), and the consumption of the n items is nonrival because there are precisely enough items to go around--one for each subject. In order to avoid free-riding and focus attention on hypothetical bias, we use a coercive payment mechanism and a private good.

In the hypothetical treatments, following previous efforts, we used passive language so subjects understood that their vote would not induce true economic consequences--i.e., no money or goods would change hands. The cheap talk treatments were also hypothetical, but included a "cheap talk" script (as described above). The language in the cheap talk script is originally from Cummings and Taylor (1999), with necessary changes due to differences in the allocation mechanism and good. In the consequential treatments, subjects were told that separate coin flips would determine: (i) which of the price levels ($5 or $10) would be utilized, and (ii) whether the corresponding votes would be economically binding. Hence, once the price level was chosen, the probability that the subjects' voting responses had real consequences was 50%. The real treatment was a straightforward referendum, but, again, since the agent was voting on the same good twice (for both $5 and $10), a coin-flip was used to determine which price level was binding, after the subjects had indicated their vote at both price levels.

A few noteworthy items should be mentioned before we proceed to the experimental results. First, all of our subjects were "ordinary" consumers (i.e., none of the experimental subjects were sports card dealers). Second, as aforementioned, subjects participated in only one treatment. Third, the experimenter was careful not to examine the votes from the first price level before asking subjects to vote in the second referendum at the second price level.

Results

A summary of the experimental data is provided in table 2. Our first order of business is to examine the field data for internal consistency. This is important since a recent study (Ariely, Loewenstein, and Prelec 2003--ALP hereafter) suggests that valuation experiments can produce results that appear coherent in the sense that subjects are responsive to within-session variation in quantities or prices, but are arbitrary in that the valuations are conditioned on design parameters that should be irrelevant to fundamental values. In one of their experiments, ALP find behavior consistent with downward-sloping demand curves when examining data associated with the same individual (a within comparison), but they found that the expressed value for common consumer products, measured with a theoretically incentive-compatible mechanism, can be considerably influenced by exposure of the subject to a clearly uninformative, random anchor.

To test for internal consistency within treatments, we utilize the binomial test (the small sample analog of McNemar's exact test for the equality of correlated proportions), the null hypothesis being that the proportion of "Yes" responses at $5 is equivalent to the proportion of "Yes" responses at $10. The alternative hypothesis is that the proportion of "Yes" responses is larger at the $5 level. At the p < 0.10 level, we reject the null hypothesis for both pricing sequences in the cheap talk and real treatments. (7) We, likewise, reject this hypothesis for pricing sequence A in the consequential treatment, but not pricing sequence B. (8) In the latter case, we find no evidence of behavior inconsistent with demand theory (i.e., a "Yes" response to $10 followed by a "No" response to $5), rather there was little response to the price change (only two subjects changed from "No" to "Yes" as the price fell from $10 to $5). We fail to reject the hypothesis of equal proportions of "Yes" responses in both pricing sequences for the hypothetical treatment. (9) In the hypothetical treatments, there were three cases where subjects exhibited behavior inconsistent with demand theory. (10) If we ignore these responses, we reject the null hypothesis. Thus, our within-subject data are generally consonant with ALP and demand theory.

We test for between-subject demand consistency by examining equality of proportions at different prices across the pricing sequences using a chi-square test. That is, we compare responses to the $5 question in sequence A to responses to the $10 question in sequence B, and vice versa, for each of the treatments. In only two cases (comparing cheap talk A:$10 and B:$5, and comparing real A:$10 and B:$5) (11) can we reject the hypothesis that these responses are equivalent. In the six other cases we fail to reject this hypothesis. (12) Visual inspection of the data confirms that the proportions of "Yes" responses associated with the $5 price level in one sequence are all higher than the proportions of "Yes" responses associated with the $10 price level from the other pricing sequence. Nonetheless, in six out of eight cases this difference is not statistically significant. We infer that demand for our PSA graded mint ticket stubs is inelastic within the price range we offered. (13)