Learning While Voting: Determinants of Collective Experimentation

This paper analyzes collective decision making when individual preferences evolve through learning.  Votes are affected by their anticipated effect on future preferences.  The analysis is conducted in a two-arm bandit model with a safe alternative and a risky alternative whose payoff distribution, or "type", varies across individuals and may be learned through experimentation.  Society is shown to experiment less than any of its members would if he could dictate future decisions, and to be systematically biased against experimentation compared to the utilitarian optimum.  Control sharing can even result in negative value of experimentation: society may shun a risky alternative even its expected payoff is higher than the safe one's.  Commitment to a fixed alternative can only increase efficiency if aggregate uncertainty is small enough.  Even when types are independent, a positive news shock for anyone raises everyone's incentive to experiment.  Ex ante preference correlation or heterogeneity reduces these inefficiencies.