Description: Reading Group: Zhigang Feng (University of Nebraska-Omaha)

Location: 360 Heady Hall

Contact Person: Juan Carlos Cordoba

Title: Optimal Taxation with Incomplete Markets—an Exploration via Reinforcement Learning

Abstract: In this paper we study the optimal taxation and government debt/asset management and price volatility in economy with incomplete markets. We develop a machine-learning-based computational method to solve the equilibrium set of endogenous state variables, along with policy and value functions. With quasi-linear preference, any finite government asset is sustainable and the Ramsey government would play a Ponzigame in absence of restriction on taxation. When taxes or assets are bounded, the asset accumulation process becomes ergodic even when the government expenditure follows a Markov process. Furthermore, we show that the celebrated Martingale convergence result in the literature is due to the weak precautionary savings motive by the government. Once the government has sufficiently strong precautionary savings motive, the ergodic property of asset accumulation process will return. Our findings indicate that the optimal taxation and government debt/asset management are intrinsically correlated with significant implications on asset pricing.