This dissertation consists of three chapters that each study one applied microeconomic theory problem. In the first chapter, I consider the problem a social planner faces in constructing a criminal justice system which addresses two needs: to protect the innocent and to punish the guilty. I characterize the socially optimal criminal justice system under various assumptions with respect to the social planner's ability to commit. In the optimal system, before a criminal investigation is initiated, all members of the community are given the opportunity to confess to having committed the crime in exchange for a smaller than socially optimal punishment, which is independent of any future evidence that might be discovered. Agents who choose not to confess might be punished once the investigation is completed if the evidence gathered is sufficiently incriminatory. In this paper's framework, leniency for confessing agents is efficient not because it saves resources or reduces risk, but because there are informational externalities to each confession. When an agent credibly confesses to be guilty, he indirectly provides the social planner additional information about the other agents: the fact that they are likely to be innocent. \par In the second chapter, I present a theory which shows how the influence of others may generate overconfidence. The argument is built on the idea that the more help an agent receives when performing a task, the less informative the score on that task will be relative to the agent's ability to perform it. Overconfident agents, who tend to benefit from more cooperation opportunities simply because they are perceived to be more skilled, will remain overconfident because the future signals they will observe will contain very little information regarding their ability. On the contrary, the scores on tasks that underconfident agents receive will be more informative, which will help them learn their true ability faster. \par Finally, in the third chapter, I compare two different systems of provision of discrete public goods: a centralized system, ruled by a benevolent dictator who has no commitment power; and an anarchic system, based on voluntary contributions, where there is no ruler. If the public good is binary, then the public good provision problem is merely an informational one. In this environment, I show that the anarchic system can always replicate any outcome of the centralized system. However, as the number of alternatives available increases, the classical free riding problem described in Samuelson (1954) emerges. As the classical free riding problem becomes more important relative to the informational free riding problem, the centralized system becomes the preferred system of the two.