Dynamic Optimal Income Taxation with Government Commitment

The optimal income taxation problem has been extensively studied in one-period models. This paper analyzes optimal income taxation when consumers work for many periods. We also analyze what information, if any, that the government learns about abilities in one period can be used in later periods to attain more redistribution than in a one-period world. When the government must commit itself to future tax schedules, intertemporal nonstationarity of tax schedules could relax the self-selection constraints and lead to Pareto improvements. The effect of nonstationarity is analogous to that of randomization in one-period models. The use of information is limited since only a single lifetime self-selection constraint for each type of consumer exists. These results hold when individuals and the government have the same discount rates. The planner can make additional use of the information when individual and social rates of time discounting differ. In this case, the limiting tax schedule is a nondistorting one if the government has a lower discount rate than individuals.