Motivated Beliefs: The Role of Uncertainty about Information Accuracy (with Stella Papadokonstantaki)
We design an experiment to study how uncertainty about information accuracy affects motivated belief formation. Subjects receive noisy signals about their IQ test performance and update their beliefs under one of three information structures: certain accuracy, where signals are correct 70% of the time; compound uncertainty, where signals are correct either 50% or 90% of the time with equal probability; and ambiguity, where signals are also correct either 50% or 90% of the time, but subjects do not know the probability of each accuracy level. To separate motivated reasoning from other biases in belief updating, subjects also update beliefs about an ego-irrelevant state, suitcase prices. Subjects update too little relative to Bayes' rule across all treatments, in both tasks, and after all types of signals. Comparing behavior across tasks, we find evidence of motivated reasoning under certain accuracy and compound uncertainty, with subjects updating their beliefs more after good news than after bad news. This asymmetry is strongest under certain accuracy, remains marginal under compound uncertainty, and is absent under ambiguity. Our results therefore provide suggestive evidence that uncertainty about information accuracy reduces motivated belief formation, yet the differences across treatments are modest.