Americans past age 60 are delaying their withdrawal from the workforce.  This trend became detectable in the 1990s, and it represents a reversal of a century-long trend toward earlier retirement.  The paper attempts to identify the groups in successive birth cohorts that have delayed their retirement in the era since the retirement age began to rise.  Do the groups which have delayed retirement earn above-average wages?  Or below-average wages?  Are workers in these groups more educated or more likely to be married and to have dependents? The analysis is performed using Current Population Survey (CPS) files, which provide detailed monthly data on the labor force status of adults in approximately 57,000 households every month.  The available monthly files cover the span from 1977 up to the present.  During the first decade of this period, the average retirement age declined; during the most recent two decades, the average age at retirement has increased. 

A crucial question about the increase in the average retirement age is the character of work effort supplied during the extra years that workers remain in the labor force.  Groups that remain longer in the workforce may choose to work either in less demanding jobs than they held during their careers or for fewer hours per week than they worked in their career jobs.  Alternatively, they may work additional years in jobs that, in terms of pay, hours, and responsibility, are very similar to their career jobs.  The monthly CPS files can be used to analyze this question.  Respondents in the survey are interviewed in 8 months over a 16-month span.  Older workers who leave a job over this period can be followed to determine whether they accept jobs in less a demanding occupation or with lower pay or work hours than their previous jobs. 

To shed greater light on the issue, we also use special CPS surveys to compare the distribution of older workers’ job tenure over time.  Since 1987 the BLS has periodically administered supplementary questions in the regular CPS to determine how long employees have worked in their jobs or for their current employers.  The answers to these questions permit us to compare the distribution of job tenures of the same birth-year cohorts at two points in time.  For birth cohorts who are near the typical ages of retirement – say, ages 60 to 67 – we can see whether the increase in average retirement age in successive birth cohorts is linked to increased tenure in jobs they held at age 55 or to an increased likelihood of working in jobs with relatively brief tenures.