Strategic Communication Games: Theory and Experiments

This dissertation focuses on the topic of information design and information exchange between economic agents theoretically and experimentally. In the first section of my dissertation, I revisit the classic problem of product information revelation using a new theoretical framework developed by Kamenica and Gentzkow (2011). In the second section of my dissertation, I design an experimental test of the predictions of Kamenica and Gentzkow (2011), an influential paper that has provided the theoretical foundations for many models in the information design literature. In the last section, I look at the problem of disclosing bias when communicating to a heterogeneous audience. A firm's incentive to provide product information is a topic of central importance in industrial organization theory. In the first section of my dissertation titled "Controlling Information to Influence Consumer Beliefs," I take a fresh look at the question of how much information firms are willing to provide consumers using the information design framework developed in Kamenica and Gentzkow (2011). In my model, I show that even without the ability to price discriminate, a firm can influence a rational consumer to change her beliefs and extract full expected consumer surplus. In contrast to the results of earlier papers such as Lewis and Sappington (1996) and Johnson and Myatt (2006), the firm chooses to provide partial information in equilibrium. In addition, I show that competition is beneficial in the sense that it forces firms to reveal information in order to differentiate themselves from one another and avoid fierce competition. Persuasion commands a huge share of economic resources and is a topic of great importance to economists. In the model of persuasion by Kamenica and Gentzkow (2011), the sender can influence the receiver's belief and therefore the receiver's action given the power to manipulate the information environment. The question of whether in reality people can learn to influence other people's prior beliefs optimally by changing the information environment as predicted by Kamenica and Gentzkow's model remains unanswered. The second section of my dissertation, titled "Bayesian Persuasion: Evidence from the Laboratory" attempts to answer this question using data generated from a series of laboratory experiments. I find that whether subjects can learn to implement the optimal persuasion strategy depends on the number of feasible strategies available to them as well as feedback about the effectiveness of each strategy. In this section, I provide data to test the theory as well as a novel experimental design that can be implemented in testing other information design theories. The final chapter of my thesis looks at the problem of communicating to a heterogeneous audience. I find that when a sender communicates with a heterogeneous audience consisting of proponents and opponents, it is beneficial for him to not disclose his bias. For example, if a conservative politician wishes to win the support of both conservatives and liberals, it is best for the politician to keep information about his political alignments to himself. Not disclosing bias improves public communication between the sender and the audience.