Speaker: John Hey, University of York
"This study reports the results of an economic experiment designed to test the Epstein and Ji (2019) model of optimal learning under ambiguity when the choice between actions can be postponed, at a per-unit-time cost, in order to observe a signal that provides information about an unknown parameter. Subjects in this experiment face a modified 2-urn Ellsberg task and may opt to receive partial information, regarding the composition of the ambiguous urn via a computerised Brownian motion. In various treatments of the experiment, we exogenously vary the signal variance and the cost of learning in an effort to investigate their effects on optimal learning. The experimental design provides data suitable for parametric estimations via appropriate econometric techniques. This kind of structural analysis provides useful statistical inference, which in turn can be used in order to test the main theoretical assumptions of the underlying ambiguity model (i.e. maxmin preferences, forward looking decision making) compared to the Bayesian baseline model and alternative specifications. In our experiment, on top of the subjects' decisions, we also collect data on their response times and basic demographics which allow us to explore potential correlations between these measures and ambiguity aversion. We find that the model’s predictions are in line with the observed choices of our subject population."