Updating beliefs with imperfect signals : experimental evidence

This article analyses belief updating when agents receive a signal that restricts the number of possible states of the world. We create an experiment on individual choice under uncertainty. In this experiment, the subject observes an urn, containing yellow and blue balls, whose composition is partially revealed. The subject has to assess the composition of the urn and form an initial belief. Then, he receives a signal that restricts the set of the possible urns from which the initial observed sample is drawn. Once again, he has to estimate the composition of the urn. Our results show that, on the whole, this type of signal increases the frequency of correct assessment. However, differences appear between validating and invalidating signals (i.e. signals that either confirm or disprove the initial belief). The later significantly increase the probability to make a correct assessment whereas validating signals reduce the frequency of correct estimations. We find evidences of lack of persistence in choice under uncertainty. The literature shows that people may persist with their choice even when they are wrong. We show that they may also change even if they are right.