We design an experiment that closely emulates and tests the standard model of complete competitive markets, without imposing parametric restrictions on preferences. Consistent with theory, aggregated elicited supply and demand curves cross at the expected dividend when there is no aggregate risk, and at a lower price when there is aggregate risk. In contradiction with theory, individual participants frequently make choices that violate ?rst order stochastic dominance. We propose a random choice model which reconciles the above mentioned ?ndings and is also consistent with additional features of the data, such as, e.g., large mistakes being less frequent than smaller ones.
TSE Working Paper, n° 17-798, avril 2017