Eliciting and aggregating individual expectations: An experimental study

In this paper we present a mechanism to elicit and aggregate dispersed information. Our mechanism relies on the aggregation of intervals elicited using an interval scoring rule. We test our mechanism by eliciting beliefs about the termination times of a stochastic process in an experimental setting. We conduct two treatments, one with high and one with low volatility.Increasing the underlying volatility affects the location of the interval, yet it does not significantly affect its length. Consequently, individuals perform significantly better in the low volatility treatment than in the high volatility treatment. Next, we construct distributions by aggregating intervals across different individuals. Our results reveal that the predictive quality of the aggregated intervals as measured by the Hellinger distance to the true distribution increases by more than 30 when increasing the aggregation level from two to eight individuals. This shows that aggregating individual intervals may be an attractive solution when market mechanisms are infeasible.