Collective experimentation: a laboratory study

We develop a simple model of collective experimentation and take it to the lab. In equilibrium, as in the recent work of Strulovici (2010), majority rule has a bias toward under experimentation, as good news for a minority of voters may lead a majority of voters to abandon a policy when each of them thinks it is likely that the policy will be passed by a future majority excluding them. We compare the behavior in the lab of groups under majority rule and under the optimal voting rule, which precludes voting in intermediate stages of the policy experiment. Surprisingly, simple majority performs better than the (theoretically) optimal voting rule. Majority rule seems to be more robust than other forms of voting when players make mistakes.