Abstract: This dissertation examines both the theoretical and empirical issues in the design of social insurance and tax programs. In each policy that is considered—social insurance, commodity taxation and targeted transfers—the public economics literature relies on an economic model that does not match the empirical evidence well. As a result, it may be misleading to derive the implications of these models for government policy, and therefore, the economics of policy reform. The broad objective of this dissertation is to develop new economic models that are consistent with the empirical evidence and that can be used to evaluate the costs and benefits of government policy. Chapter 1 of this dissertation examines the optimal level of social insurance by developing a theoretical framework that incorporates the decision to take up social benefits. This framework highlights that several behavioral parameters—ones that have heretofore been ignored in the literature on optimal social insurance design—that are key to assessing the economics costs and benefits of social insurance policy. These parameters are estimated using a new empirical research design. Chapter 2 of this dissertation (joint with Rai Chetty and Adam Looney) presents empirical evidence showing that the salience of a commodity tax affects behavioral responses to taxation. We propose a simple bounded rationality model to explain why salience matters, and show that it matches several additional stylized facts. In the model, agents incur second-order (small) utility losses from ignoring some taxes, even though these taxes have first-order (large) effects on social welfare and government revenue. Using this framework, we derive elasticity-based formulas for the efficiency cost and incidence of commodity taxes when agents do not optimize fully. Chapter 3 of this dissertation focuses on the optimal intra-household allocation of resources. It provides the conditions under which the government is able to implement an equitable allocation across spouses in the family.