The Timing of Retirement: a Comparison of Expectations and Realizations

Abstract

In this paper, I employ data drawn from the Social Security Administration's Retirement History Survey (RHS) to study the accuracy of expectations concerning the timing of retirement. The RHS is ideally suited for this purpose, in that it collects information on retirement plans, and follows respondents through time so that one can identify actual dates of retirement. The data are consistent with the view that, when asked to report an expected date of retirement, individuals name the most likely date (i.e. a mode, rather than a mean). Furthermore, these forecasts are highly accurate. There is very little evidence that individuals' expectations were systematically biased during periods in which Congress legislated large real increases in social security benefits. This suggests either that the benefit increaser were anticipated, or that unanticipated changes in benefits have little effect on retirement. The paper also describes differences in the accuracy of expectations by population subgroup.