“Multidimensional Information and Rational Inattention” by Ms. Keri P. Hu
Ms. Peicong (Keri) Hu
Ph.D. Candidate in Economics
University of California San Diego
This paper studies a rational inattention model in which a principal communicates with an uninformed agent about correlated multidimensional information. The principal agrees with the agent about appropriate actions to be taken given the states, though the players have potentially different preferences about relative importance of each dimension. I show that if the agent’s loss from suboptimal actions in one of the dimensions could be influenced by the principal, it would benefit the principal to increase the weight that agent places on the loss of this dimension when agent’s cost of processing information is either low or high, regardless of the principal’s weight on the loss. If cost is in the intermediate range, extra incentive provision on the dimension important to the agent could be harmful, particularly at the lowest cost such that decision quality of the other dimension would be affected. In this case, I characterize the condition under which it benefits the principal to provide (only) information about the dimension that is more important from her perspective but withhold information about the dimension more important to the agent. Withholding information of either dimension will not favor the principal when the agent’s cost is low or high. I find that the principal could be better off if the dimension important to the agent involves more complex information, but always gets 3; weakly lower payoff if the dimension important to the principal has more information. The model has implications for controlling incentives and information available in organizations.