Anonymous Procedures for Condorcet's Model: Robustness, Nonmonotonicity, and Optimality

Abstract

In Condorcet's model of information aggregation, a group of people decides among two alternatives a and b, with each person getting an independent bit of evidence about which alternative is objectively superior. I consider anonymous procedures, in which the group's decision depends only on the number of people who report a or b, not their identities. A procedure is called incentive compatible for a person if she wants to report truthfully given that others report truthfully. I show that if an anonymous procedure is incentive compatible for both a person who is significantly biased toward a and a person who is significantly biased toward b, then it is incentive compatible for any person, regardless of his preferences and prior beliefs; also, if it is not trivial, it must be nonmonotonic, with an additional report for a sometimes decreasing the probability the group chooses a. I define the "supermajority penalty" (SP) procedure and show that when there are significant biases in both directions, the SP procedure is the optimal anonymous incentive compatible procedure from the point of view of an unbiased person.

Central banks’ forecasts and their bias: Evidence, effects and explanation

“…Austen-Smith and Banks, 1996, Ladha, 1992, Chwe, 2010. Our work also relates to that of Dietrich and List (2004) who present a model in which jurors make decision based on a common body of evidence rather than the state of the world.…”

Mechanisms without transfers for fully biased agents

“…Bouton et al (2018) argue that the unanimity rule is also inferior to a majority rule with veto power when agents sometimes have a private preference against one of the two options. Chwe et al (2010) shows that adding conflicts of interest between the agents can imply that non-monotonic voting rules are optimal. Ali and Bohren (2019) study the question whether a principal can benefit from banning deliberations between agents and find that doing so can be helpful if the voting rule is non-monotonic.…”