Uncertainty, choices and prices in a market for lotteries

We consider an experimental setting where agents receive one stylized piece of information at a time about the value of a lottery. We find that Knightian uncertainty about the prior distribution of true lottery values does not hamper decision making by agents and markets. On a mean squared error criterion, Bayesian updating is closer than simple averaging in predicting market prices and individual bids and offers, even in treatments with uncertainty where Bayesian updating should not be feasible given the limited information set. Bayesian updating also outperforms adaptive expectations in relation to market prices