Transitional Dynamics and Long-run Optimal Taxation Under Incomplete Markets

By Ömer Tuğrul Açıkgöz

http://d.repec.org/n?u=RePEc:pra:mprapa:50160&r=dge

Aiyagari (1995) showed that long-run optimal fiscal policy features a positive tax rate on capital income in Bewley-type economies with heterogeneous agents and incomplete markets. However, determining the magnitude of the optimal capital income tax rate was considered to be prohibitively difficult due to the need to compute the optimal tax rates along the transition path. This paper shows that, in this class of models, long-run optimal fiscal policy and the corresponding allocation can be studied independently of the initial conditions and the transition path. Numerical methods based on this finding are used on a model calibrated to the U.S. economy. I find that the observed average capital income tax rate in the U.S. is too high, the average labor income tax rate and the debt-to-GDP ratio are too low, compared to the long-run optimal levels. The implications of these findings for the existing literature on the optimal quantity of debt and constrained efficiency are also discussed.

The results of this paper will upset people across the political spectrum. First, the public debt to GDP ratio should be much higher than currently in the US. This is because the public debt allows households to overcome their borrowing constraints, thus the best is for the government to borrow up to the natural borrowing limit (which is not the debt ceiling). Second, labor income taxes should be higher, because this finances the debt and reduces the volatility of household income. Third, capital income taxes should be lower, as this favors the accumulation of precautionary savings. The paper also highlights that policies that are welfare-maximizing in the long run can leads to significantly dominated outcomes in the short-run. This also shows that as so often in the optimal tax literature, optimal policy is difficult to find and results can easily be reversed by changing some aspect of the model. The future will tell whether this analysis will be robust.