Association for Public Policy Analysis & Management

This paper extends prior research that investigates the distribution of time in retirement. The length of retirement has expanded and in combination with changes in sources of income such as the shift towards defined contribution pensions and contributory savings in Individual Retirement Accounts has implications for post-retirement wellbeing. Because full lifespans are not readily observed in current data on individuals, this requires the simulation of the events between the last observation in the data and individuals’ deaths. In the analysis, we apply hot-deck imputation methods to Health and Retirement Study data to construct a synthetic sample of people who lived natural human lifespans to measure the duration of time between their withdrawal from the labor force and death. This allows us to use those data to examine not only time spent in retirement and how it varies across groups but also the characteristics of peoples’ lives such as adequacy of financial resources. We also make use of CPS and SIPP data to provide information on historic patterns of the gap between retirement and death across groups to provide further context for the analysis.

We find that Blacks and women, but not white men, with low educational attainment have shorter retirements and spend larger shares of their retirement time with morbidities (measured by needing assistance with activities of daily living). Defined benefit pensions help men with low socioeconomic status retire relatively young when they are available and reduce the share of retirement time impaired. Because of changes in pension design and coverage we find that some groups with shorter than average life spans lack pensions. They also tend to hold smaller amounts of other assets. Due to this, they cannot entirely compensate for their relatively shorter lives by retiring early. Due to the increased reliance of Americans on defined contribution plans and the influence of this change in pension design on retirement timing, the distribution of time spent in retirement is likely to become more unequally distributed in the future. The estimates suggest that retirement time, especially for men, is becoming more unequal by SES.