Optimal Retirement Policies with Present-Biased Agents

Abstract

Abstract This paper incorporates quasi–hyperbolic discounting into a Mirrlees taxation model to study the design of retirement policies for present-biased agents. I show that the government can improve the screening of productivity by exploiting time inconsistency. This is done by providing commitment to sophisticated agents and taking advantage of the incorrect beliefs of naïve agents. This can be achieved even if the degrees of present bias and sophistication are private information. I also demonstrate how the government can implement the optimal mechanism using retirement savings accounts and social security benefits.