Henrik Kleven (Princeton) and Owen M. Zidar (Princeton) have a new NBER working paper, “Automation and Optimal Taxation: A Task-Based Theory” (Sept. 2026). Here is the abstract:

We characterize optimal labor and capital income taxation in a task-based model of automation. Workers and machines are perfect substitutes in automatable tasks, which run from the bottom of the skill distribution up to a threshold. Workers supply labor on the extensive margin. Capitalists supply machines at a finite elasticity. In this general-equilibrium automation economy, optimal tax formulas take canonical partial-equilibrium forms. Yet automation substantially changes optimal tax rates because wages and capital returns are endogenous. We calibrate the model to the US wage distribution and automation exposure. In equilibrium, the middle class is the most automated, with machine intensity peaking around the 40th wage percentile. Relative to a no-automation benchmark, the optimal labor tax schedule is more progressive: a larger EITC subsidy at the bottom, lower taxes in the middle, and higher taxes at the top. The optimal capital tax is sizable but unaffected by automation.