Optimal Taxation with Time-Inconsistent Agents

Abstract      

People with time-inconsistent preferences tend to make intertemporal choices different from their original intentions. In particular, time-inconsistent agents make sub-optimal saving decisions. This paper studies the optimal savings and insurance policy in an economy with time-inconsistent agents who privately observe their skill. I introduce a mechanism that can elicit private information at zero cost by threatening sophisticated time-inconsistent agents with o_-equilibrium path policies that can undo commitments, and fooling naively time-inconsistent agents with empty promises. The government can implement the full information efficient optimum, which is better than the constrained optimum obtained with traditional proposals to increase savings, like linear savings subsidies or mandatory savings rules. I introduce regressive savings subsidies for naive agents and government loans with multiple repayment plans for sophisticated agents as policy instruments for decentralization. I show that welfare increases monotonically with the population of time-inconsistent agents. In essence, the presence of time-inconsistent agents improves the government's ability to provide insurance.