Voting and Experimentation

This paper analyzes a dynamic voting model where individual preferences evolve through experimentation. Individual votes reflect not only current preferences but also the anticipated effect of elected alternatives on future preferences and votes. The analysis is conducted in a two-arm bandit model, with a safe (status quo) alternative and a risky alternative whose payoff distribution, or type”, varies across individuals and may be learned through experimentation. Under any voting rule, society experiments less than any individual would if he could dictate future decisions, due to a control-loss effect. Depending on the nature of uncertainty, majority-based experimentation also has a systematic bias compared to the utilitarian policy. For large groups with independently distributed types, this control-loss effect annihilates the value of experimentation, prompting individuals to vote myopically. However, even with independently distributed types, a positive news shock for anyone raises everyone's value function and incentive to experiment. Efficiency increases with ex ante preference correlation. The paper also discusses the effect on experimentation of the ability to commit and of asymmetric information.

For more information, contact Andrew Postlewaite.