« Back to Results

Optimal Policies in a Behavioral World

Paper Session

Friday, Jan. 5, 2018 10:15 AM - 12:15 PM

Marriott Philadelphia Downtown, Liberty Ballroom Salon A
Hosted By: American Economic Association
  • Chair: Christian Moser, Columbia University

Optimal Taxation with Behavioral Agents

Emmanuel Farhi
Harvard University
Xavier Gabaix
Harvard University


This paper develops a theory of optimal taxation with behavioral agents. We use a general
behavioral framework that encompasses a wide range of behavioral biases such as misperceptions,
internalities and mental accounting. We revisit the three pillars of optimal taxation:
Ramsey (linear commodity taxation to raise revenues and redistribute), Pigou (linear commodity
taxation to correct externalities) and Mirrlees (nonlinear income taxation). We show
how the canonical optimal tax formulas are modified and lead to a rich set of novel economic
insights. We also show how to incorporate nudges in the optimal taxation frameworks, and
jointly characterize optimal taxes and nudges. We explore the Diamond-Mirrlees productive
efficiency result and the Atkinson-Stiglitz uniform commodity taxation proposition, and find
that they are more likely to fail with behavioral agents.

Optimal Paternalistic Savings Policies

Christian Moser
Columbia University
Pedro Olea


We study optimal savings policies when there is a dual concern about under-saving for retirement and income inequality. In our model, agents differ in time preferences and earnings ability, both unobservable to a planner with paternalistic and redistributive motives. We characterize the solution to this two-dimensional screening problem and provide a decentralization using realistic policy instruments: forced savings at low incomes---similar to Social Security---but a choice between savings accounts with different subsidies and caps at high incomes---like 401(k) and IRA accounts in the US. Offering more choice in savings for high-income individuals acts as a screening device that facilitates redistribution. We calibrate our model to microdata on income and wealth accumulation and find that the current US savings and tax system is off the Pareto frontier, with large welfare gains available from simple reforms.

Optimal Income Taxation with Present Bias

Benjamin B. Lockwood
University of Pennsylvania


Work often requires up-front costs in exchange for delayed benefits—one must search for a job before becoming employed, paychecks are typically delayed by a few weeks, and a promotion may come only after months or years of extra effort—and mounting evidence documents present bias over labor supply in the face of such delays. This paper studies the implications of such present bias for the optimal income tax schedule. I derive expressions for optimal tax rates as a function of observable elasticities and present bias, conditional on income. Present bias lowers optimal marginal tax rates, with a larger effect when the elasticity of taxable income is high. If labor commitment contracts are feasible, tax rates depend on the residual uncorrected degree of present bias. Residual bias can arise either because workers are naive or because commitment contracts may be infeasible due to limited liability constraints which prevent firms from imposing fines on workers who quit. I calibrate the model using both existing estimates of present bias and a new estimate of residual present bias using subjective well-being trends following US welfare reforms in the 1990s. All evidence suggests bias is concentrated at low incomes. Numerical simulations show that for modest redistributive preferences, optimal marginal tax rates are substantially negative across low incomes, comparable to those under the Earned Income Tax Credit in the US. Yet the model also generates novel results about optimal tax timing, with implications for improving the schedule of EITC payments.

Optimal Retirement Policies with Time-inconsistent Agents

Pei Cheng Yu
University of New South Wales


This paper develops a general theory for the design of retirement policies, like social security and retirement accounts, within a Mirrlees taxation framework with hidden present bias and sophistication. The paper shows how policies can utilize the time inconsistency of agents to improve welfare above the constrained efficient optimum. In particular, in an environment with both time-consistent and time-inconsistent agents, welfare increases monotonically with the population of time-inconsistent agents. For implementation, the paper focuses on the design of social security and retirement accounts. The optimal policy has social security benefits decreasing in progressivity with the initial withdrawal age. It also allows early withdrawals from retirement accounts only when there are large income discrepancies. The coexistence of both policies screens sophistication and present bias. These proposals outperform traditional policies, like linear savings subsidies or mandatory savings, by increasing redistribution and output efficiency. The resulting welfare improvement could be quantitatively significant depending on the size of the time-inconsistent population.

Optimal Illiquidity

John Beshears
Harvard Business School
James Choi
Yale University
Christopher Harris
University of Cambridge
David Laibson
Harvard University
Brigitte Madrian
Harvard University


We calculate the socially optimal level of illiquidity in a stylized retirement savings
system. We solve the planner’s problem in an economy in which time-inconsistent households
face a tradeoff between commitment and flexibility (Amador, Werning and Angeletos,
2006). We assume that the planner can set up multiple accounts for households: a
perfectly liquid account and/or partially illiquid retirement savings accounts with early
withdrawal penalties. Revenue from penalties is collected by the government and redistributed
through the tax system. We solve for the socially optimal values of these
penalties, and the socially optimal allocations to these accounts. When agents have heterogeneous
present-biased preferences, the socially optimal system has three accounts: (i)
a fully liquid account, (ii) an account with an early withdrawal penalty of approximately 100%, and (iii)
an account with an early withdrawal penalty of approximately 10%. With heterogeneous preferences,
the socially optimal retirement savings system in our stylized model looks surprisingly
like the existing U.S. system: (i) a liquid account, (ii) an illiquid Social Security account
(and defined benefit pensions), and (iii) a 401(k)/IRA account with a 10% penalty. The
socially optimal allocations to these accounts and the predicted equilibrium flows of early
withdrawals---“leakage”---also match the U.S. system.
Joshua Schwartzstein
Harvard Business School
Stefanie Stantcheva
Harvard University
Simone Galperti
University of California-San Diego
Roozbeh Hosseini
University of Georgia
Dmitry Taubinsky
University of California-Berkeley
JEL Classifications
  • E0 - General
  • H2 - Taxation, Subsidies, and Revenue