UC San Diego

In 1996 the United States passed a major reform of welfare, the federal and state program that had provided and continues to provide money to poor, mostly single mothers since 1935. One element of the reform was that it capped benefit receipt at no more than five years regardless the age of a recipient's child. This dissertation focuses on understanding what led to these time-limits. Chapter 1 makes the case that changes in the welfare caseload-- brought on by population changes, the Supreme Court, and other government programs--could have led to its reform. The problem was that too many people receiving welfare were people for whom it was not intended. The solution Washington chose was to cut the program's generosity intertemporally by instituting five-year time limits. Chapter 2 adds rigor to the welfare reform story in Chapter 1 by analyzing (1) how lack of observability affects the design of a social insurance program, and (2) how it interacts with an increasing population who find going on social insurance attractive. It shows that a lack of observability affects the generosity and incentives of the program because what is preferred with observability often cannot be implemented when it is absent. Finally, it shows that a rising population of the poor can induce programs to switch from allowing the poor not to work to ones that do require them to work because the high types subsidizing the program slash the benefit to reduce excessive redistribution. Chapter 3 examines our ability to use observable information to explain future welfare usage and finds that it has gone down since reform. I argue that this result is consistent with the social insurance function of welfare being more important than it was before reform. I note, however, that it is also consistent with a rise in the importance of unexplained variation