Leakage from Retirement Savings Accounts in the U.S.

Abstract

This paper generates new, aggregate estimates of retirement savings flows in the U.S. from 2003 to 2015, and provides detailed estimates of leakage from tax-preferred retirement savings accounts to pre-retirement age individuals. We create a nationally-representative panel of individuals using a sample of administrative tax data with over 140 million person-year observations. These data contain information on retirement contributions, distributions, and transfers between accounts. We estimate that between 2003 and 2015 distributions from defined contribution plans and IRAs to individuals age 50 or younger were equal to 22 percent of the contributions made by this age group. When estimating the correlation between common life events and the probability of leakage, we find that job separations correspond with an increase in the probability of leakage of over 200 percent. Job separations generating the receipt of unemployment insurance (UI) -- a proxy for an involuntary job separation -- are associated with higher leakage than non-UI separations. Other types of events, such as income shocks, home purchases, and the onset of tuition payments are also associated with leakage.

Keywords: Leakage, Retirement, Tax Policy, Defined Contribution, Defined Benefit, Job Loss, Tax Data

JEL Classification: H24, H31, J14, J63