Essays on information, inattention, and ambiguity

This dissertation consists of three essays studying economic agents with non-standard reactions to information. The standard model is often inadequate because it permits neither inattention nor ambiguity aversion. This dissertation provides both pure and applied theoretical analyses of these two phenomena. The first essay models an agent who has a limited capacity to pay attention to information and thus conditions her actions on a coarsening of the available information. An optimally inattentive agent chooses both her coarsening and her actions by constrained maximization of an underlying subjective expected utility preference relation. The main result axiomatically characterizes the conditional choices of actions by an agent that are necessary and sufficient for her behavior to be seen as if it is the result of optimal inattention. The agent's utility index, cognitive constraint and prior are uniquely identified. The second essay analyzes the implications of advertising in a model where consumers are optimally inattentive. Firms compete by choosing both prices and advertising levels. Consumers easily observe price but have a limited capacity to pay attention to information about quality. Advertisements increase consumer capacity for attention. An increase in capacity for attention results in more information processed by each consumer, which raises the likelihood that a high quality good is purchased but leads to an increased price. An exogenous decrease in the cost of advertising has a positive impact on equilibrium price but an ambiguous effect on equilibrium profit and surplus. Advertising generates some effects documented in the literature through a single mechanism. The third essay studies strategic voting when voters have pure common values but exhibit Ellsberg-type behavior as modeled by maxmin expected utility preferences. The Condorcet Jury Theorem states that given subjective expected utility maximization and common values, the equilibrium probability that the correct candidate wins goes to one as the size of the electorate goes to infinity. In contrast, this essay provides sufficient conditions so that the equilibrium probability of the correct candidate winning the election is bounded above by one half in at least one state. As a consequence, there is no equilibrium in which information aggregates