Optimal Dynamic Nonlinear Income Taxation under Loose Commitment

This paper examines an infinite-horizon model of dynamic nonlinear income taxation in which there exists a small probability that the government cannot commit to its future tax policy. In this "loose commitment" environment, we find that even a little uncertainty over whether the government can commit yields substantial effects on the optimal dynamic nonlinear income tax system. Under an empirically plausible parameterization, numerical simulations show that high-skill individuals must be subsidized in the short run, despite the government's redistributive objective, unless the probability of commitment is higher than 98%. Loose commitment also reverses the short-run welfare effects of changes in most model parameters. In particular, all individuals are worse-off, rather than better-off, in the short run when the proportion of high-skill individuals in the economy increases. Finally, our main findings remain qualitatively robust to a setting in which loose commitment is modelled as a Markov switching process.