Retirement and Wealth

Abstract

This paper analyzes the relationship between retirement and wealth. In a simple model where the only heterogeneity is in leisure preference, other things the same, those who retire early accumulate more wealth while still working, enabling them to support themselves over their longer retirement period. Moreover, characteristics that encourage earlier retirement also encourage additional saving. If there were heterogeneity in both leisure and time preference, however, this simple relation is broken. Early retirees do not necessarily save more. Using data from the first four waves of the longitudinal Health and Retirement Study, a cohort of individuals born from 1931 to 1941, we estimate reduced form retirement and wealth equations. Linked employer provided pension plan descriptions and social security administrative records are central to the analysis. The value of the pension and social security beyond current period accrual is measured by the difference between the present value of the benefit stream resulting from additional work until the date of retirement and the present value of a stream of benefits equal each year to the value of benefit accrual in the initial period. This measure, which we call the premium value, captures any excess value from the spikes at early and normal retirement age in a defined benefit plan. But it also has zero value in the case of a defined contribution plan. Calculating benefit increments on the assumption that benefits are claimed as soon as eligible after retiring, and that respondents link delayed benefit claiming with delayed retirement, the estimated retirement equation indicates that a higher future reward from pensions and social security encourages postponed retirement. Factors leading to early retirement do not systematically generate higher saving. Many independent variables do not have symmetric effects in the retirement and wealth equations. Unobservables from the retirement and wealth equations are only weakly correlated. A related finding, not easily reconciled with a simple life cycle model of saving, is that higher pension wealth and social security wealth do not substitute for other forms of wealth, but add to total wealth. In addition, other findings support a more complicated view of the underlying behavior. Most importantly, despite a significant payoff to waiting, retirees do not time the acceptance of their social security benefits so as to maximize expected value. Most respondents take their social security benefits as soon as eligible after retirement. This raises questions about the way social security and pensions are calculated as explanatory variables in reduced form retirement equations. These and other findings, e.g., on measuring retirement and on the role of partial retirement, raise doubts about the value of using reduced form retirement equations to estimate the effects of changing such social security policies as the early retirement age. Reduced form retirement equations must be used with great caution in situations where they are analyzing new policy initiatives. Unobserved heterogeneity interacts with observable variables to produce the estimated coefficients in these equations, but these interactions are not necessarily the same if the policy changes in new ways.