Using Consumption Data to Derive Optimal Top Income and Capital Tax Rates

Two of the most influential results in optimal tax theory — Saez’s (2001) formula for the top income tax rate and Atkinson and Stiglitz’s (1976) case for not taxing savings — are usually invoked separately, and this paper shows they are two sides of one coin: the revenue-maximising top income tax equals Saez’s τ if and only if uniform commodity taxation applies and the optimal savings tax is zero, so any departure from that joint benchmark forces a trade-off between the two instruments. The reason the split matters empirically is that consumption is much less concentrated than income at the top: the authors’ own estimates from the 2005–2021 waves of the PSID put the Pareto coefficient of income at 1.8 but of consumption at about 3.1, and plugging the consumption tail into the same formula would cut the implied top income tax from 74% to 57%. Working in a Mirrleesian model where agents work, consume and save for retirement, the paper derives representations of optimal top income and savings taxes in terms of three consumption-based sufficient statistics — the ratio of Pareto coefficients, the ratio of compensated elasticities, and a scaling parameter identified by the elasticity of intertemporal substitution — and shows that consumption data, not savings data, are what identify how the combined wedge splits. Calibrating to available empirical estimates, the paper finds it optimal across all its calibrations to shift part of the top earners’ burden from income onto savings, with an optimal savings tax between 12.5% and 30% and a top marginal income tax falling from 74.1% to between 63% and 70.4%; the paper is explicit that the size of the shift is sensitive to the calibration strategy and that identifying the pass-through elasticity separately from the ratio of Pareto tails is critical, since varying the two across their plausible ranges moves the optimal savings tax from 4.5% to 26.3%. It also draws a pointed implication for policy: because the two benchmarks stand or fall together, Diamond and Saez’s (2011) simultaneous case for high top income taxes and for positive capital taxes is, on the paper’s argument, internally inconsistent — and it is the second recommendation the paper’s evidence supports.

Christian Hellwig | Macro Paper Warehouse

Using Consumption Data to Derive Optimal Top Income and Capital Tax Rates