KD2 Lab, Team Rooms A&B

Strategizing under Rule and Vote Uncertainty: An Experiment

I n a controlled laboratory experiment, we examine voting behavior under rule uncer[1]tainty, i.e., uncertainty about the voting rule itself. We compare behavior under three voting-rule conditions: simple plurality (1R), plurality with runoff (2R), and their proba[1]bilistic mixture (1R/2R) that is a lottery generating either 1R with known probability p, or 2R with probability 1−p. Following the previous literature, we conjecture that 1R/2R raises computational complexity and thus mitigates strategic manipulation. We test dif[1]ferent models – either heuristic-based or rational – of (i) the formation of beliefs about other voters’ behavior, and of (ii) the resulting voting decisions. We find that beliefs tend to be formed in a myopic manner in all experimental conditions. With repetition, however, the accuracy of the belief formation process improves and we observe convergence between the beliefs about votes and the actual votes. For voting, the model with highest (resp., lowest) predictive power is strategic (resp., sincere) voting, with some variation across conditions. We further show that in all experimental conditions – 1R, 2R, and 1R/2R – voting behavior can be well described by a simple heuristic: best-responding to (myopic) beliefs under the 1R rule. In particular, such combination of myopic belief formation and bounded best-responding is an efficient way to strategize under rule uncertainty. Overall, our initial conjecture is not supported by the experimental data. Rule uncertainty steers the voters neither towards sincerity§ If anything, it contributes to promoting strategic behavior.