[EM] Proposed experiment

[Jameson Quinn]

I'd be interested in running an behavioral-economics-type experiment on
voting behavior. I imagine a game matrix of 9 voters and 3 candidates, with
each candidate having a known payout for each player. For each condition,
we'd have a separate group of experimental subjects. We'd run two
non-binding "pre-election polls" and one "election", and then pay real money
for the payout. I could put up the money, create the experimental protocol,
and program a web app to run the elections, but I'd like help actually
getting diverse sets of subjects. I imagine three payoff matrices (positive,
zero, and negative sum; see below) and up to six voting systems (approval,
range, MJ, SODA, condorcet for negative sum, and IRV for positive sum), for
a total of 11 conditions. The payoff matrices would be as follows:
 Group Size Candidate
Scenario 1 (zero-sum) A B C
 a 4 4 1 0
 b 2 0 3 2
 c 3 0 2 4
 Total payout 16 16 16

Scenario 2 (positive-sum CW) A B C
 a 4 3 1 0
 b 2 0 3 1.5
 c 3 0 2 3
 Total payout 12 16 12

Scenario 3 (negative-sum CW) A B C
 a 4 4 0.5 0
 b 2 0 3 2
 c 3 0 1 4
 Total payout 16 11 16


These scenarios present both a chicken dilemma between candidates B and C,
and a contrast between a strong (for positive sum) or weak (for negative
sum) Condorcet winner (candidate B).

To save money and subjects, the four voters of voter group A could be
represented by 2 actual experimental subjects with double-weighted votes.
Thus, the total subjects necessary for a full set of 14 experimental
conditions would be 98. Ideally, we'd run at least one scenario for each of
the voting methods twice; thus we could also use up to 140 subjects. For 98
subjects, the total payout would be at most 234 monetary units, but probably
under 200. Here in Guatemala, for a 30-minute experiment like this, those
monetary units could probably be 1-2 quetzales ($0.13-$0.26); in the US,
they would have to be $1-2, for a total payout cost of up to $400. Again,
I'd put up that money.

If you're interested, we'd have to talk about creative ways to get subjects
and space. The best would be if there were someone here who is an
undergraduate (at an undergrad-focused college) or a graduate student (at a
university), who could get access to a pool of psych-101 student guinea
pigs. Obviously, even then, getting a hundred subjects is not just a matter
of asking for them; but with some coordination, I think we could manage it.

Once the experiment was done, we could write it into a paper, including also
mathematical discussion and proofs, simulations, and historical analysis;
and I'm confident that we could get published. Yes, the statistics would be
weak, perhaps too weak to really discern behavioral differences between
Approval, Range, and MJ even if such differences exist; but I'm sure that no
matter what happens, the results would advance our knowledge. I don't care
whether I am counted as the primary author on that paper.

Jameson
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20111021/98e04238/attachment.htm>