Economic Theory Seminar - Francisco da Silva (Deakin University)

Title:  Communication through biased intermediators

Abstract: We study biased intermediated communication between a sender and a receiver.  We prove that the information that can be transmitted from the sender to the receiver is exactly the same as with direct  (non-mediated) communication, provided there are (at least) two intermediators who do not communicate with each other.  In that sense, the sender is not harmed by not being able to communicate with the receiver directly.  This is the case even if the sender does not know the agents’ biases and regardless of the rules of communication (cheap talk, bayesian persuasion, etc.).  We discuss the implications of our results to a related information design problem, where a decision maker designs the statistical experiment to be performed by a possibly biased agent.  We show that, if the decision maker is able to manipulate the data without causing any statistical loss, she can implement her preferred statistical experiment despite the agent being biased.  If,  however,  manipulating  the  data  has  a  statistical  cost,  then  strategies that reduce the overall informativeness of the experiment (like reducing the sample) might actually be (second-best) optimal for the decision maker, because they may require less costly manipulation in order to dissuade biased agents from misreporting.  We discuss these ideas in the context of medical research.