Retirement Analysis, Social Security and Pensions
Gustman, Alan L., Steinmeier, Thomas L., NBER Reporter
Our research efforts over the past decade have focused on three central issues in the economics of aging: retirement, Social Security, and pensions. We have examined how retirement is defined and have contributed explanations for the wide differences in retirement behavior among individuals. We have investigated the variety of incentives observed in pension plans and the sharp trends in these incentives over time. We also have considered related public policy questions pertaining to Social Security, pension regulation, and retirement income policies.
In addressing these topics, other important behavioral and institutional questions arise. How do fixed working hours affect labor supply and retirement choice? What is the nature of the long-term employment relationship and what are the consequences for pensions, retirement, and other dimensions of labor market outcomes? How well informed are individuals about Social Security and their own pensions, and how does misinformation affect retirement behavior?
Much of our work revolves around a structural retirement model. This model incorporates our findings on the proper specification of the opportunities governing the choice of retirement outcomes, and it has been used to analyze a variety of public policies.
Consider first the question of how to define retirement. Our work suggests that it is not fully informative, and sometimes is misleading, to treat retirement as an either/or choice. Retirement behavior is most properly treated as involving labor market flow among at least three states, including partial retirement. Transition rates into and out of these states determine the stock of full-time workers, partially retired, and fully retired.
Theory suggests that if individuals could choose their hours in a job, most would gradually reduce work hours as they grow old. However, the majority of workers proceed directly from full-time work to retirement. The reason is: most jobs do not have flexible hours. In order to work part time, most workers must change jobs. Our work shows that these hours constraints are pervasive. Most individuals are not free to retire on their main job by smoothly reducing their hours of work from full time to zero, while remaining with their employer. As a result, they face what is effectively an all-or-nothing derision about retirement from that job. In order to retire partially, they must take a new job and forgo the specific training accumulated over a long period on the main job. Consequently, partial retirement jobs offer much lower wages than can be earned on the main job.
Not only are constraints on hours common, but they are important in estimating retirement models. A retirement analysis that ignores hours constraints could suggest that workers are less subject to the influence of retirement incentives and retirement programs than is in fact the case. We also emphasize that the response of partial retirement to incentives may be substantially different from the response of full retirement to those same incentives. Studies that assume the two effects are identical may misinterpret the relationship of retirement to pension and Social Security incentives.
A central focus of our work has been to estimate a structural retirement model that allows for these features of the labor market, and to apply that model to related policy analysis. The model includes different wage offers for full-time work and for partial retirement work, as well as incentives from both Social Security and pensions. In a recent paper, we also have included the analogous incentives created by retiree health insurance. We also adjust the analysis to reflect the effects of the income tax. We have estimated the model using data from the Retirement History Study (RHS) and the National Longitudinal Study (NLS) of Mature Women.
Our simulations indicate that the major peaks of retirement are a result of the incentives created by pension plans, and by discontinuities in the Social Security benefit formula. …