Learning under compound risk vs. learning under ambiguity – an experiment

We design and conduct an economic experiment to investigate the learning process of agents under compound risk and under ambiguity. We gather data for subjects choosing between lotteries involving risky and ambiguous urns. Agents make decisions in conjunction with a sequence of random draws with replacement, allowing us to estimate the agents’ beliefs at different moments in time. For each type of urn, we estimate a behavioral model for which the standard Bayesian updating model is a particular case. Our findings suggest an important difference in updating behavior between risky and ambiguous environments. Specifically, even after controlling for the initial prior, we find that when learning under ambiguity, subjects significantly overweight the new signal, while when learning under compound risk, subjects are essentially Bayesian.