Using the CPS to track retirement trends among older men.
It has always been difficult to identify the age at which people retire because separation from the labor force is often neither abrupt--part-time work is very common among older workers--nor final--many older persons reenter the labor force after a period of absence. In addition, retirement status is best defined by current work activity for some purposes, while for others, pension receipt is the more appropriate criterion. Given the types of data that are most readily available, a simple definition of retires is often used, such as those who receive Social Security retirement benefits, or those above a certain age, such as 5, who are not in the labor force.
Transitions from work to retirement are probably best tracked by longitudinal surveys, which follow the same individuals for a period of time. Among the most notable of these are the Retirement History Survey and the Continuous Work History Sample of the Social Security Administration, and the National Longitudinal Survey, conducted by the Center for Human Resource Research, Ohio State University. Longitudinal surveys are particularly useful because of the considerable amount of demographic and other personal information available on individuals in the survey. A drawback of many longitudinal surveys is that they focus on persons in a limited age range at the time of the initial survey, which means that they cannot provide comparisons between these and other cohorts of workers.
One does not need to follow the same people to track a group's labor force trends. Unlike the longitudinal surveys, the Current Population Survey (CPS) relies on a rotating sample--that is, a household (technically, an address) is in the sample for a limited time and is then replaced. In the CPS, 25 percent of the sample changes each month. But, while the survey does not follow the same people for long periods, the sample can "represent" the same group over time. In other words, within the limits of sampling reliability, any random sample of persons 55 years of age at one point in time would represent the same group as a different sample of 54-year-olds surveyed a year earlier.
Because of the long history of the CPS and the frequency of observation, the survey can provide an excellent overview of changes in retirement trends. The data can be used in three ways. The cross-sectional view examines the labor force characteristics of persons of different ages at a fixed point in time. The time-series view examines the behavior of one or more demographic groups at different times. A third, the cohort view, follows the same people, or a sample representing the same people, as they age. This view has the advantage of permitting one to consider the unique history of each population group when assessing its present labor force status.
"Retirement" data from the CPS have generally been used with the time-series approach to track changes in labor force participation rates for broad age groups, usually persons 55 to 64 years and 65 years and over. However, since 1963 CPS data have been available on labor force characteristics by single year of age and by sex, for persons age 55 to 74. Thus, the CPS provides a better vantage point than most longitudinal surveys in that it follows work histories of many cohorts through their older years.
This summary presents these previously unpublished data for older men and estimates of rough retirement histories for different generations of these men. A simple definition of retirement is used for this purpose; all men over age 55 who are not in the labor force are deemed to be retired. Conversely, all who are working, whether full or part time, and all those actively looking for work are not retired.
Labor force participation rate--the proportion of the population in the labor force at each age--for men between ages 55 and 74 are shown in table 1 for the years 1963-83. From these estimates, two types of retirement histories are calculated, using the cohort perspective, for the 1904-22 birth cohorts. (Insufficient data are available for earlier cohorts, and later cohorts are not old enough to be included.) Table 2 shows the proportion of the population of each cohort that had retired at any particular age. These estimates are additive, that is, adding across gives the proportion of a cohort that had retired as of a certain age. These retirement rates are depicted to chart 1, which shows the percentage of men in even-year birth cohorts who were out the labor force as of selected ages. The heights of the five sections of each bar represent the percentages of men who were retired by age 61, and of those who subsequently retired at ages 62, 63 and 64, 65, and 66 to 70. Of course, the retirement histories of the younger cohorts are not yet complete.
The second type of retirement history is provided in table 3, which gives the probability of someone who is in the labor force as of a certain age leaving the labor force the next year. For example, this table shows the probability that someone who was in the labor force at age 65 in 1970 would be out of the labor force at age 66 in 1971.
The difference in the two types of "retirement rates" is that the first shows the proportion of the population of each cohort leaving the labor force at each age, while the second shows the proportion of those in the labor force at each age leaving it the next year. In other words, table 2 answers the question, "At what age did men in each cohort leave the labor force?" For example, among the 1904 cohort, 3.1 percent left the labor force at age 60, and 2.2 percent did so at age 61. Table 3 answers the question, "What is the probability of someone who was in the labor force as of a certain age retiring (that is, leaving the labor force) the next year?" Among the 1904 cohort, 3.5 percent of 59-year-old labor force participants retired at age 60; of those left in the labor force, 2.6 percent retired at age 61, and so forth.
In using any of these data, one should keep in mind that, as in any sample industry, the results shown may differ from the true population values, largely because of sampling error. The problem of statistical reliability of the estimates becomes more acute as the size of the group being counted declines. Thus, apparently inconsistent trends or odd occurrences (such as the two positive retirement rates shown in tables 2 and 3) may be attributable, at least in part, to sampling error, and to other types of measurement error such as response or coding errors. Users should intrepret the estimates for specific cells in each table with some caution; the data are best used to show general trends in retirement behavior.
|Printer friendly Cite/link Email Feedback|
|Author:||Rones, Philip L.|
|Publication:||Monthly Labor Review|
|Date:||Feb 1, 1985|
|Previous Article:||One-fourth of the adult labor force are college graduates.|
|Next Article:||Women and minorities: their proportions grow in the professional work force.|