A Model of State Aggregation

Optimizing over all of the states of the world together might be difficult even for a machine. This paper adds to the behavioral literature by introducing a model, in which the agent aggregates the states together, even though she is aware of the entire state space. As a result of the state aggregation, the person solves several problems with fewer variables instead of the initial problem with the entire state space. When the person is SEU-maximizer, the decisions are not affected by the way the states get aggregated. In our model people still have subjective priors over states and events, however, they lump some states together in a non-linear way, which leads to different choices. The paper provides axioms for a state aggregation model, discusses identification of the state aggregation from choices in a complete market setting, discusses comparative statics due to changes in the state aggregation and demonstrates how the model explains a number of ambiguity paradoxes.