Learning and information

Psychologists have found several common biases when individuals are asked to assess probability. In most instances individuals will rely too heavily on prior experience especially when incoming information is hard to process. When prior experience may be hard to interpret or remember, newer information may weigh too heavily into decision-making. Further, individuals tend to discount low probability events as impossible. I propose a model of Bayesian updating to account for these effects based on the psychological costs of updating. I call this model the Limited Learning Model.; The Limited Learning Model can account for all commonly documented violations of expected utility theory in laboratory experiments. Further it provides a plausible explanation for why these violations might occur. If individuals bias beliefs toward prior beliefs when new information may be difficult to understand, then we should expect to see the systematic violations of expected utility theory that are commonly observed. I compare this theory of decision-making under uncertainty to the leading models using theoretical analysis and empirical techniques.; An example of the Limited Learning Model is used to demonstrate the ability of this model to describe observed behavior in laboratory experiments. Further, using mathematical proof, I show that in order to avoid undesirable implications, any model of decision under uncertainty must have a form similar to the Limited Learning Model.; Using a large and comprehensive experimental data set I show that the Limited Learning Model has performed as well as the leading models in terms of econometric prediction. In particular, the Limited Learning Model performs better than all but the rank-dependent model when compared using standard techniques.; Beyond predictive ability and theoretical appeal, the Limited Learning Model has substantive implications for policy and firm behavior. An application to crop insurance is shown to predict behavior previously observed, but not explained by previous theory. An application to agricultural contracts is also proposed to explain deviations from theoretically optimal contracts observed in the processing tomatoes industry.