Optimal Illiquidity in the Retirement Savings System

John Beshears, James Choi, Christopher Clayton, Christopher Harris, David Laibson, Brigitte Madrian

NBER Retirement Research Center Paper No. NB 14-05
Issued in September 2014

This paper calculates the socially optimal level of illiquidity in a retirement savings system. We study an environment in which time-inconsistent agents face a tradeoff between commitment and flexibility (cf. Amador, Werning and Angeletos 2006). For this analysis, we assume that the agent has access to two accounts: a perfectly liquid account and an illiquid retirement savings account with an early withdrawal penalty (0% ≤ π ≤ 100%). When agents have homogeneous present-biased preferences (with short-run discount factor β), we find that the socially optimal retirement savings account should have a penalty that is approximately equal to the level of present bias, 1 − β. In this case, the penalty roughly offsets the present bias of the representative agent. For example, if β = 0.7, then the socially optimal early withdrawal penalty rate is approximately 30%. However, when agents have heterogeneous preferences, with a range of β values, we find that optimal policy disproportionately addresses the needs of low β agents. In an illustrative calibration with β values distributed uniformly between 0.1 and 1 (with a mean β value of 0.55), we find that the optimal savings system is characterized by a retirement savings account that is essentially perfectly illiquid (i.e., with an early withdrawal penalty rate of π ≅ 100%). In other words, our analysis with heterogeneous preferences suggests that savings should be divided between two accounts: one account that is completely liquid and one account that is completely illiquid (like a defined benefit pension plan).

PDF (965 K)Executive Summary (62 K)