The problem of the divided majority: Preference aggregation under uncertainty

- I experimentally study a novel repeated-game, divided majority setting.
- I compare three voting mechanisms under two information structures.
- Without knowing others' preferences Plurality Voting leads to inefficient outcomes.
- Approval Voting and Borda Count dissolve information imperfections efficiently.
- The Condorcet loser is often defeated independently of the treatment.

This paper reports on the results of a series of experimental laboratory elections. The novelty of the design allows me to study the extent to which voting methods defeat the Condorcet loser and elect the Condorcet winner in a repeated-game, divided majority setting. I assess and compare the performance of three voting mechanisms, Approval Voting, Borda Count, and Plurality Voting under two information structures. Voters either know the preference structure in the electorate or hold no information regarding other voters' preferences. With enough experience, the number of elections won by the Condorcet loser is fairly low across voting methods and information structures. Approval Voting and Borda Count dissolve information imperfections efficiently and allow voters to implement the Condorcet winner, independently of the underlying information structure. The frequency with which the Condorcet winner is elected under Plurality Voting crucially depends on available information. When voters are uninformed about the preference structure in the electorate, Plurality Voting fails to implement the Condorcet winner. This is costly and decreases total welfare.