Liquidity in a Socially Optimal Retirement Savings System


Brigitte Madrian, James Choi, John Beshears, and David Laibson


Tax-deferred retirement savings accounts in the U.S. (e.g., IRA, 401k, 403(b), Thrift Savings Plan, etc.) are partially illiquid. Withdrawals before age 59½ incur a 10% tax penalty (in addition to income taxes). The 10% early withdrawal penalty was originally legislated in 1974. At that time, the penalty was a source of contention, with the Senate preferring a much higher penalty than the House. To our knowledge, neither economists nor policy makers have engaged in a systematic effort to evaluate the social optimality of the liquidity of the IRA/401k system.

The liquidity of the current system has consequences. For every $1 contributed to the accounts of savers under age 55, $0.45 simultaneously flows out of the 401k/IRA system, not counting loans (Argento et al., 2012). Much of this leakage occurs for socially desirable reasons, like providing liquidity during periods of financial hardship.

We seek to evaluate the optimality of the illiquidity features in the U.S. retirement savings system using a behavioral model that includes both spending shocks (e.g., health costs and other sources of financial hardship) and self-control problems (e.g., dynamically inconsistent preferences like those introduced by Strotz, 1955). The integrated model is based on the mechanism design framework developed by Amador, Werning, and Angeletos (2006) and extended by Beshears et al. (2013).