Divided majority and information aggregation : theory and experiment

We propose a theory-based experimental approach to compare the properties of approval voting (AV) with those of plurality. This comparison is motivated by the theoretical predictions that, in our aggregate uncertainty setup, AV should produce close to first-best outcomes, while plurality will not. The experiment shows, first, that welfare gains are substantial. Second, both aggregate and individual responses are in line with theoretical predictions, and thus with strategic voting. Finally, subjects' behavior under AV highlights the need to study equilibria in asymmetric strategies.