Elicitation of expectations using Colonel Blotto

We develop a mechanism based on the Colonel Blotto game to elicit (subjective) expectations in a group-based manner. In this game, two players allocate resources over possible future events. A fixed prize is awarded based on the amounts the players allocate to the realizedevent. We consider two payoff variations: under the proportional-prize rule, the award is split proportionally to the resources that players allocate to the realizedevent; under the winner-takes-allrule, the full award is given to the player who allocate the most resources to the realizedevent. When probabilities by which events realizeare common knowledgeto the players, both games are Bayesian–Nash incentive compatiblein the sense that (expected) equilibrium allocations perfectly reflect the true realizationprobabilities. By means of a laboratory experiment, we find that in a setting where realizationprobabilities are common knowledgethe game with the proportional-prize rule (Prop) elicits better distributions compared to both the winner-takes-allvariation (Win) and a benchmark mechanism based on an individual-based proper scoring rule(Ind). Without common knowledgeof realizationprobabilities Prop is at least as good as Ind, showing that it is possible to use a game to elicit expectations in a similar fashion to using a proper scoring rule.