Experimental Design Details
Our experiment will represent a common value election between two alternatives, A and B, where the best option for every voter (or experimental subject) will be the same, but where the alternative that best serves voters' interest will not be known to every voter. To determine voters' best alternative, an equally likely state of the world (also A or B) will be independently drawn.
The electorate will be composed of 5 (human) subjects who must collectively choose an alternative, where the committee's decision will be made through a simple majority election, i.e. whichever color receives more votes will be the collective decision that is applied to everyone; ties will be broken by a fair coin toss. To represent the behavior of voters whose preferences are independent of the state of the world, we also consider two additional votes cast randomly by the computer, which we refer to as the partisan bias. The partisan bias will have three possible values that correspond to two votes for A, two votes for B, or one vote for each, A and B. Each of these events are equally likely. The partisan bias in our experiment will provide a hurdle that subjects must cross to achieve their best alternative. The state of the world and the level of the partisan bias will be randomly determined prior to subjects making voting decisions.
Voters will not know the state of the world ex ante and will not be told what the correct decision is until after the collective decision has been made. Voters will have identical preferences that depend only on the group decision and the state of the world. If the collective decision matches the state of the world, then all participants will receive a high payoff of 110 experimental points, Otherwise, all participants will receive a low payoff of 10 experimental points.
Before voting, some voters will be exogenously assigned as expert voters, and will be perfectly informed of the state of the world, which tells them which alternative best serves the common interest. In addition, an expert voter will also be told the value of the partisan bias. To determine whether a voter is assigned expertise, in each election the computer will generate a random INFORMATION NUMBER for each subject from a uniform distribution between 1 and 100 points. Subjects will become experts based on an exogenously assigned cutoff over INFORMATION NUMBERS that is symmetric, commonly known, and fixed in each treatment. If the computer randomly generates an INFORMATION NUMBER that is higher than the exogenously assigned threshold for each treatment, the subject will not be told the state of the world or the partisan bias (and is not charged the expertise fee). If the INFORMATION NUMBER is lower than the exogenously assigned threshold, then the state of the world and partisan bias is privately revealed. There is no cost for the information (political expertise). We administer expertise in this way because our method is relatively simple for subjects to understand and calculate the level of expertise. As a critical part of our design, expert voters do not know exactly about the number of how many voters are experts, but all the voters know the distribution of expert voters, and can form a posterior belief regarding the number of expert voters.
After expertise has been administered, each subject will be asked to make a voting decision. We design a novel method to elicit an individual's willingness to vote that takes advantage of the Becker-Degroot-Marshak (BDM) mechanism. Once a subject reports her willingness to vote, the computer will independently generate a voting cost from a uniform distribution between 1 and 100. If a voter had reported a vote choice, and if her individual voting cost is lower than her reported willingness to vote, then the subject will pay the voting cost, and her vote is included in the collective decision. Instead, if a subject's individual voting cost is higher than her reported willingness to vote, then the subject's vote will not be counted toward the committee's decision and she will not be charged the cost of voting.