Tailored Bayesian Mechanisms: Experimental Evidence from Two-Stage Voting Games

Abstract

Optimal voting rules have to be adjusted to the underlying distribution of preferences. However, in practice there usually is no social planner who can perform this task. This paper shows that the introduction of a stage at which agents may themselves choose voting rules according to which they decide in a second stage may increase the sum of individuals’ payoffs if players are not all completely selfish. We run three closely related experimental treatments (plus two control treatments) to understand how privately informed individuals decide when they choose voting rules and when they vote. Efficiency concerns play an important role on the rule choice stage whereas selfish behavior seems to dominate at the voting stage. Accordingly, in an asymmetric setting groups that can choose a voting rule do better than those who decide with a given simple majority voting rule.

Keywords: Bayesian voting experiments, revelation principle, Two-stage voting

JEL Classification: C91, D70, D82