Eliciting Multiple Choices: a Theoretical Analysis of Incentives in Experiments

Experiments elicit choices from a given set of decision problems. When multiple decision problems are offered in the same experiment, however, behavior in one decision problem may be distorted by the choices made in others. In our terminology, this is a violation of incentive compatibility. We show that, when preferences respect stochastic dominance, the Random Decision Selection (RDS) mechanism that pays for one randomly-selected decision is essentially the only incentive compatible mechanism. If an experiment pays for multiple decisions, then a ‘no complementarities’ condition must be assumed. We summarize the implications for experimental design in Section II.