Mitigating floods and pestilence: examining the provision of public goods under uncertainty

Abstract

This dissertation contains two essays on the provision of a specific type of public good, specifically public goods that affect the probability with which different states of the world occur. Two potential examples of this type of public good are levees and monitoring networks for disease. To distinguish this type of public good from those that have no relationship with risk we adopt the term risk-modifying public good. The first essay focuses on the provision of risk-modifying public goods within a mechanism design framework. The main result of the first essay is proof that there exists a family of social choice functions that provide a Pareto efficient level of a risk-modifying public good and that these social choice functions are implementable in dominant strategies. The family of social choice functions we identify provide a unique level of the public good and vary only in how the gains from providing the public good are distributed amongst the agents in the economy. The second essay is empirical in nature and focuses on measuring the returns to information gathered by a monitoring network for an invasive wind borne crop disease (soybean rust). What makes this information a risk-modifying public good is that it is used by agricultural producers to inform themselves about their risk of being affected by this invasive species and to make fungicide application decisions. In this type of situation, applying fungicides when the disease is actually present can be thought of as one state-of-the-world and applying fungicides when one shouldn't another. Thus, a state-of-the-world can be thought of as a combination of agent's beliefs about the environment relative to the reality that they face. A dynamic model of farmer decision making is presented where the information from the monitoring network can help farmers better learn about their risk of soybean rust infection and can potentially be used as an alternative to a farmer scouting soybean fields if the information is sufficiently accurate. Using this model, the marginal returns produced by additional years of monitoring and the question of the optimal spatial arrangement of sentinel plots are considered.