This paper examines a dynamic model of nonlinear income taxation in which the government cannot commit to its future tax policy, and individuals are quasi-hyperbolic discounters who cannot commit to future consumption plans. The government has both paternalistic and redistributive objectives, and therefore uses its taxation powers to maximize a utilitarian social welfare function that reflects individuals' true (long-run) preferences. Under first-best taxation, quasi-hyperbolic discounting exerts no effect on the level of social welfare attainable. Under second-best taxation, quasi-hyperbolic discounting increases (resp. decreases) the level of social welfare attainable when separating (resp. pooling) taxation is optimal. In stark contrast to previous studies, this result implies that some individuals can actually be better-off in the long run as a result of their short-run impatience.

This paper addresses the question as to whether it is optimal to use separating or pooling nonlinear income taxation, or to use linear income taxation, when the government cannot commit to its future tax policy. We consider both two- period and inÖnite-horizon settings. Under empirically plausible parameter values, separating income taxation is optimal in the two-period model, whereas linear income taxation is optimal when the time horizon is inÖnite. The welfare e§ects of varying the discount rate, the degree of wage inequality, and the population of high-skill workers are also explored. For realistic changes in these parameters, separating income taxation remains optimal in the two-period formulation, and linear income taxation remains optimal in the inÖnite-horizon model.