Inattention and Belief Polarization

Disagreement persists over issues that have objective truths. In the presence of increasing amounts of data, such disagreement should vanish, but it is nonetheless observable. This paper studies persistent disagreement in a model where Bayesian rational agents learn about an unobserved state through noisy signals. We show that agents (i) choose signal structures that are more likely to reinforce their prior beliefs and (ii) choose signal precisions that are inversely related to the precision of their prior beliefs. We call the former the confirmation effect and the latter the complacency effect. The complacency effect may lead agents to stop updating their beliefs entirely, leading to permanent disagreement among agents about the true state of nature. Taken together, the confirmation and complacency effect imply that the beliefs of ex ante identical agents over time tend to cluster in two distinct groups near the boundaries of the belief space. The complacency effect is stronger and more general when information cost is proportional to channel capacity, compared to when it is proportional to reduction in entropy. We argue that in some settings, it may be more natural to model the cost of attention as proportional to channel capacity, rather than the common practice of specifying information cost as proportional to entropy reduction.