Traditional agency models study communication in settings where agents are better informed than principals. However, in many settings, principals have better information than agents. In this dissertation, I explore the question of how communication occurs between informed principals and uninformed agents. I study this question in the context of two settings.

In the first setting, which is explored in Chapter 1, top management in a firm (principal) has better information about some economic parameter (e.g. consumer demand or preferences), which influences an employee's (agent) performance on a project. In the second setting, which is studied in Chapter 2 and coauthored with Alexandra Lilge, top management has better information about the employee's promotion prospects within the firm.

Across both settings, I show that, in contrast to conventional wisdom, the principal can benefit from not disclosing private information. In the first setting, nondisclosure allows the principal to obtain better contractual terms with the agent in a moral-hazard context. In the second setting, nondisclosure benefits the principal by curtailing the agent's incentive to engage in accounting misreporting.