Experimental Design
In this design, each of the two treatments (B and C in Plott and Sunder, 1988) will be compared. Treatment B can thus be seen as our control group and C as our treatment. However, our analysis does not rest on a direct comparison of treatments as these market experiments were used to test the prediction of rational expectation theory according to which prices should reflect true asset value. Thus, part of our analysis is simply to check for each treatment (B and C) whether prices converge to true value.
I. Instructions and Procedures
We will use the same instructions and terminology from Plott and Sunder 1982 with the appropriate modifications as noted in Plott and Sunder (1988) (i.e., “The only change was from two to three states of the world.”).
Market Procedures
Oral double auction markets will be conducted. Bids and asks will be written on the whiteboard in the manner described in Plott and Sunder 1982 (and further detailed in Plott 1991). To mitigate any transcription errors, these bids and asks will be simultaneously entered into a computer software program. Space permitting, all bids and asks will be left on the whiteboard until the market period ends. An auctioneer, who will be unaware of the research agenda, will be trained to conduct these auctions independently from the researchers. The auctioneer will have two assistants: one will write orders on the board, while the other will enter the orders into the computer. The oral double auction market will be organized as follows:
Bids and offers will still be tendered verbally using cardboards, which is an important distinction between oral and computerized markets (see Williams 1980). Once the auctioneer acknowledges a subject’s verbal bid (offer), the auctioneer’s first [second] assistant will write [enter] the bid (offer) and the index of the agent onto [into] the whiteboard [computer]. The bid (offer) is repeated verbally by the auctioneer and the floor is then open for new bids and offers. Only the last bid and offer are standing and they remain standing until accepted, cancelled or replaced. Acceptance can occur at any time. Consistent with Plott 1991 (p. 31) once a trade occurs, the book will be cleared. This will be represented by a horizontal line beneath the bid and ask. All bids and asks above the line are no longer valid.
Any buyer (seller) is free at any time during the period, to raise his hand and make a verbal bid (offer) to buy (sell) one unit of the asset at a specified price. The bid (offer) must be higher (lower) than the outstanding bid (offer) should one exist. Any seller (buyer) is free at any time to accept or not accept the bid (offer) of any buyer (seller). If a bid (offer) is accepted, a binding contract has been closed for a single unit and the buyer and seller will record the contract price to be included in their earnings. Any ties in bids or acceptances will be resolved by a random choice of buyer or seller by the auctioneer. Except for the bids (offers) and their acceptance subjects are not to speak to one another. Following the original design closely, we will use an open book. Also, we will allow traders to borrow money as in the original design. Subjects will also record their transactions as in Plott and Sunder (1988). A calculator will be available to facilitate the calculations.
When asked a question such as "What am I supposed to do?" the auctioneer will reread the relevant portion of the instructions: "The experimenters do not care whether or how you participate so long as you stay within the confines of the rules." (see Plott 1982).
II. Markets to be Replicated
As noted above, we intend to replicate Markets 4, 5, 7, 8, and 9. We will run four experimental sessions for each market (totaling 20 sessions). We plan to follow the design of each market as described in Table 1 of Plott and Sunder (1988) with respect to the number of periods, number of traders, initial endowment, and dividend amounts and probabilities.
Markets 5 and 8 in Plott and Sunder (1988) were populated with experienced subjects from Markets 3 (or 4) and 7, respectively. We will follow the original paper's approach by using the same subjects in our four Market 5 (8) experiments as in our four market 4 (7) experiments. That is, we will recruit 96 subjects for a two-day experiment. Half of these subjects will participate in our Market 4 and Market 5 replications, while the others will participate in our Market 7 and 8 replications. For the Market 7 and 8 replications we will keep the cohort of 12 subjects together in both experiments. That is, the same 12 subjects that participated in Market 7 together will also participate in Market 8 together. With respect to the Market 4 and 5 replications, we will recruit subjects in a systematic manner to ensure that the twelve subjects involved in a given ‘inexperienced’ market will all not participate (together) in the same subsequent ‘experienced’ market (similar to the Market 5 recruitment in the original paper, as the subjects came from Markets 3 and 4).
III. Analyses
We will conduct the same analyses as in Plott and Sunder (1988). Specifically, we will reproduce Tables III, IV, and V and Figures 4, 5, 7, 8, and 9 for each session of each market. We will consider this a successful replication if our data show the same (or stronger) support for the Rational Expectations model that was observed in the original set of experiments. That is, the replication is successful if Rational Expectations significantly outperforms the Prior Information and Maximin models in predicting observed prices, allocations, and profits, according to Wilcoxon tests on various measures of the deviation from the theoretically predicted quantities.