Essays on Information Design and Regulation

Abstract

My dissertation focuses on information design problem and regulation. In the first chapter, I study how a ban on price discrimination affects firms' investment decisions and social welfare in an intermediate good market. When price discrimination is allowed, an upstream firm can extract all profits of downstream firms by using discriminatory pricing. Thus downstream firms will not invest knowing that all benefits from the investment will be captured by the upstream firm. If instead price discrimination is banned, the upstream firm is forced to yield some benefits from investment to a more efficient downstream firm. I show that this can result in a strict improvement of social welfare. Researchers may be tempted to manipulate the reporting of experimental data. This is likely to affect which experiments are conducted, how the reported results are interpreted and welfare. The second chapter addresses this question in a setting where a sender has limited ability to commit to the choice of information structure, or experiment, and a receiver observes both the chosen experiment and the reported results. This gives rise to a novel mode of communication: a mixture of cheap talk (no commitment or full manipulation) and Bayesian persuasion (full commitment or no manipulation). I show that there is a clear Pareto-ranking among three different modes of communication: both the sender and the receiver are strictly better off as the communication environment changes from cheap talk through communication with partial commitment to Bayesian persuasion. This strict welfare ranking holds for any level of conflict of interests between the sender and the receiver. In the last chapter I study a principal-agent problem in which the principal is an information user and the agent is an information provider. The agent can conduct an experiment that reveals information about the unknown true state. The agent has private information about which experiments are feasible, his type. While the principal can observe both the experiment conducted by the agent and the experimental results, she cannot observe the type of the agent. First I note that there are some cases in which the principal can achieve the first-best outcome despite the information asymmetry. When the first-best outcome is not achievable, there are two kinds of optimal decision mechanisms: (1) one which assigns the best experiment to each type with the distortion in the ex post optimal decisions and (2) the other which achieves the ex post optimal decisions by not assigning the best experiments. I provide sufficient conditions that determine which decision rule is optimal.