skip to content

Keynes Fund


Summary of Project Plan

The last 25 years has seen the development of a large and extremely influential public finance literature that derives simple, applicable principles for tax design from a variant of the static Mirrlees income tax model.1

Analytically, this ‘sufficient statistics’ literature made progress by moving away from the traditional Mirrleesian mechanism design approach (optimal allocations, subject to incentive compatibility) and instead studying a ‘dual’ tax design problem, in which the government chooses tax schedules directly. An important limitation, however, has been the relative lack of attention given to dynamic considerations – in particular, how best to tax savings in a multi- period economy. This is the main focus of my project.

The analysis of tax design in multi-period dynamic economies is challenging, because of the large number of potential behavioural interactions across the different time periods. Treatments of savings taxes in the sufficient statistics literature have, to date, abstracted from this problem either by limiting to two-period models, or focusing exclusively on long-run steady-state outcomes.2

Yet alongside this there exists a significant theoretical literature that analyses multi-period versions of the original Mirrlees problem in its ‘mechanism design’ form. This literature has been criticised for yielding insufficient practical lessons for tax design.3 I show that this assessment may have been too hasty. Taking an off-the-shelf infinite-horizon model of unobservable consumption taste shocks, I pioneer the equivalent of the ‘dual’ representation that has been so influential in static settings. Crucially, I show that just one cross-period elasticity effect matters for understanding optimal taxation in t. This simplification arises as a direct consequence of the separable structure of demand over time that is assumed in the dynamic Mirrlees problem. Thus, simple lessons for dynamic tax policy remain possible, by exploiting the mechanism design approach.

This simplification opens the door to first-order qualitative conclusions about the desirable structure of dynamic taxes. In particular, I show that in the model I study, it is generally optimal to set a positive marginal tax rate on savings – a previously unknown conclusion. In future work, I plan to test the sensitivity of this to alternative environments with information asymmetries – in particular, a more traditional Mirrleesian model with productivity shocks.

1 Saez (2001, REStud) was the most significant initial contribution, building on a special case treated in Diamond (1998, AER). Piketty and Saez (2013) in the Handbook of Public Economics provide a clear overview, whilst Diamond and Saez (2011, Journal of Ec. Perspectives) directly expound the novel policy lessons this literature has provided.

2 Diamond and Saez (2011, Journal of Ec. Perspectives) discuss savings taxes in the two-period setting; Piketty and Saez (2013, Econometrica) introduced the steady state approach. Stantcheva (2020, NBER Annual Review) surveys alternative approaches.

3 See, for example, the discussion in Stantcheva (2020, NBER Annual Review)



Charles Brendon


Charles Brendon is the El-Erian Fellow in Macroeconomics at the Faculty of Economics, University of Cambridge. His research interest is in Macroeconomic Theory and the Economics of Optimal Taxation.


Share this Project