Dynamic income taxation, redistribution, and the ratchet effect

In this paper we extend the classic optimum linear income tax model to a twice repeated game. A utilitarian policy-maker decides on the second-period tax system on the basis of observed first-period behaviour. We summarise our findings as follows. First, the static optimal linear tax results cannot be expected to hold in a dynamic setting. Second, more productive households are likely to mimic the behaviour of less productive households. Third, the optimal dynamic first-period tax rate might be negative; then income is redistributed from low- to high-income households. Fourth, if the utility function is well-behaved, then the optimal dynamic first-period tax rate is lower than the optimal static tax rate. Fifth, if the optimal static tax rate separates, then the dynamic and the static optimal tax rates coincide. Sixth, if a zero tax rate separates, then the optimal dynamic tax rate is positive.