Revised worklife tables reflect 1979-80 experience.
The Bureau of Labor Statistics has been producing worklife estimates for the U.S. population since 1950. Initially, these estimates portrayed workers as being continuously active from the time of initial labor force entry until final retirement. In 1982, after completing a major study of worklife methodology, the BLS published its first set of increment-decrement, or multistate, working life tables for the years 1970 and 1977. Based on observed rates of labor force entry and exit at all ages, those tables for the first time quantified the impact of midlife labor force withdrawal and reentry on worklife duration. Their publication drew responses from many economists involved in litigation of wrongful injury or death cases. Several such responses have been published in the Monthly Labor Review, and some of the refinements proposed by readers have since been implemented in BLS worklife research.
This analysis incorporates some of those refinements, updates the 1982 study, and presents a new set of official worklife estimates based on patterns observed during the period 1979-80. It also adds two new dimensions to the discussion, for the first time exploring how race and educational background affect lifetime labor force behavior.
Method of the new study
As was the case with previous BLS worklife estimates, the new figures have been calcualted from information collected in the Current Population Survey (CPS), a nationwide monthly household survey conducted by the Bureau of the Census on behalf of the BLS. Individuals are interviewed during each of 4 successive months, and again in the same 4 months of the following year. Questions focus on the labor force behavior of household members during the week preceding each interview.
For the period of study, CPS records have been matched so that each person's status at the beginning and end of a 12-month interval can be compared. Labor force transitions have been registered if labor force status changed between the two reference dates. Transition rates have been developed for each age, sex, race, and educational category to identify the group's unique pattern of labor force mobility.
The worklife tables for 1977 were derived from a single matched sample of about 40,000 persons, interviewed in January 1977 and again in January 1978. To provide the additional demographic detail presented below, the curent study pools six matched samples focusing on individuals' labor force status in a given month of 1979 and in the same month of 1980. Specifically, the study focuses on persons interviewed in the following months of each of the 2 years: January, March, May, July, September, and November. Together the six samples include nearly 255,000 matched responses.
The multistate working life table model is extremely sensitive to rapid changes in rates of labor force entry or withdrawal. Tables based on a recessionary period, during which labor force exits increase, present a very bleak picture of lifetime labor force involvement. Conversely, those calculated during periods of rapid recovery or expansion tend to overstate the average degree of lifetime labor force attachment. To avoid the problems caused by the cyclical swings of the early 1980's, the current study rests on data for a somewhat earlier but less turbulent period, 1979 to 1980.
The cost of avoiding cyclical irregularities in this way is that certain secular trends may be understated. To the extent that underlying patterns of male and feamle labor force involvement have converged since 1980, the sex differentials in this report may overstate those now in evidence. However, until it is possible to update the tables again, the 1979-80 period has been judged the most viable for calculation of multistate worklife estiamtes.
Factors affecting worklife duration
In the working life tables for 1970 and 1977, worklife duration was treated as if it were a simple function of sex and age. Tables were prepared separately for men and women, giving no additional demographic or functional detail by race, educational attainment, occupation, or other characteristics that might distinguish high from low turnover groups.
In reality, labor force attachments are influenced by a variety of factors, including training, health, marital and family responsibilities, economic opportunity, and additional sources of income. However, it is not feasible to control for all of these factors in computing worklife expectancy. For example, while worklife estimates by occupation are in particular demand, it would require development of a clustering scheme for occupations by prevailing work patterns, together with study of job changes among potentially hundreds of occupations, to computer them. The only other approach is to assume that no such changes occur. Because neither of these alternatives is practical, no such estiamtes are computed. Nonetheless, this tudy does add two new dimensions to the estimation of worklife: race and education. Tables are presented separately for each of these two variables. However, the combined impact of race and education has not been computed because the present matched sample is too limited to develop reliable joint probabilities.
Working life tables show the combined effects of mortality and labor force mobility rates on lifetime labor force involvement. The mortality estimates used in this report are averages of the 1979 and 1980 values released by the National Center for Health Statistics. Tables by race incorporte the effects of sex- and race-specific mortality. Those focusing on education employ only sex-specific rates, because there are no comparabble mortality tables by education. Of course, access to health care is apt to be correlated with schooling. If it were possible to quantify this relationship, the tbles would probably show still wider discrepancies between the worklives of the less and more educated.
Apart from the factors listed above, all of which affect the behavior of workers, certain properties of the data may also influence our perception of that behavior. Model assumptions and sample design are two such factors. The BLS worklife model has changed little since 1977; it should cause no marked discontinuities. However, the expanded sample, in which subsets are observed at six different points during the year, captures more labor force mobility than was evident in the earlier tables. In particular, the new sample includes two groups of persons whose labor force behavior was observed, retrospectively, in May and July of each of the 2 years. This is the period during which students and seasonal workers are most likely to report themselves as economically active.
Neither worklife expectancies nor net flows appear to have changed greatly between 1977 and the end of the decade. but rates of labor force accession and separation rose noticeably. Because of modification of the sample, such differences should not be interpreted as an accurate reflection of "changes" in mobility rates.
Developments between 1977 and 1979-80
The general relationships observed in earlier worklife tables remained valid through the end of the 1970's. Women continued to have higher probabilities of labor force exit and reentry than men. Consequently, men continued to have longer worklives, on average, than women. (See table 1.) Not surprisingly, the worklife expectancy of persons in the labor force was higher than that of the inactive population. The gap was small for young persons, but widened considerably with age. Men who were in the labor force at age 50 could expect to work 6.3 years longer than other men at that age. The comparable figure for women was 4.6 years.
Between 1977 and 1980, the cross-sectional participation rates of men changed very little. (See table 2.) Those of older teenagers and men above the age of 55 dropped slightly. In contrast, the activity rates of women continued to climb. There was an overall gain of more than 3 percentage points, with the largest change occurring in the age range 25 to 54. This change in cross-sectional rates signaled shifts in the underlying patterns of labor force involvement. However, because the multistate model builds on flow data (that is, entry and exit rates) rather than stocks (activity rates), the relationship between changes in activity rates and worklife values is sometimes weak.
During the period in question, the observed participation rate for men 16 and older edged downward from 77.7 percent to 77.4 percent, while their worklife expectancy rose by .6 years. Worklife expectancies held steady for men aged 55 to 64, despite a modest drop in activity rates. Further, despite the observed drop in participation rates of those 65 and older, worklife expectancies for these men actually rose slightly as life expectancy increased.
Among women 16 and older, whose total activity rate rose by 3.1 percentage points, worklife duration increased by 1.8 years. The fact that expectancies rose across the board indicates that women of all ages were developing a stronger bond with the job market.
The relationship between lifespans and worklife expectancies is particularly revealing. (See table 3.) Between 1977 and the end of the decade, the life expectancy of the average 20-year-old man rose by half a year. His worklife expectancy went unchanged, the entire gain being allocated to nonmarket activity. Women of the same age also gained a half year of life, but allocated this additional time to labor force activity and reduced nonmarket time by an average of .7 years, for a total worklife gain of 1.2 years. As a result, the sex differential in worklife continued to narrow. Whereas in 1977 the 20-year-old woman could expect to work 70.7 percent as long as her male counterpart, by 1979-80 the ratio had risen to 73.9 percent.
The trend toward earlier retirmenet observed between 1970 and 1977 appeared to have leveled off in the closing years of the decade. The worklife expectancy of 65-year-old men, which had dropped from 3.1 years in 1970 to 1.9 in 1977, was 2.3 years by the end of the decade. (See table 4.) For women of a comparable age, the figure had dropped from 1.4 to 1.1 years, but stood at 1.5 years by 1979-80. The model's insensitivity to hours of work makes it difficult to interpret these changes. They may well reflect the impermanence of many retirement decisions, and the fact that so-called retirees often resume part-time jobs for either economic or social reasons.
The new tables show little change in the proportion of persons expected to die while economically active. (See table 4.) In 1977, the figures for men and women were 27.0 percent and 9.5 percent, respectively, compared with 27.4 percent and 10.4 percent for 1979-80.
Differentials by race and education
Although expansion of the data base for the present study has obscured our view of changing labor force mobility rates, this loss has been more than offset by an improved perspective on racial and educational differentials. Data users have long pressed for more focused tables, and the new estimates should meet some of their more urgent needs.
Life table models derive their estimates of lifetime behavior not from panel stuides but from a series of cross-sectional surveys collected during a single year. Each age group in the population being analyzed contributes a single year of life to the synthetic whole. It is possible to dervice group-specific estimates only if the group is closed to entry and exit. If its members remain so classified for life, the experiences of older persons can be used to derivce a synthetic "future" for the young.
In the new tables, the population is subdivided by sex, race, and educational attainment. While subject to misclassification, each of these traits is normally fixed during the adult years. Sex and race are particularly stable, and beyond the mid 20's, education--especially as classified here--is also relatively fixed. Only persons who already have some advanced training are likely to continue schooling, and attainment levels, once achieved, cannot be lost. Because these groupings are closed, they satisfy the constraints of the model. And because they relate closely to labor force behavior, they are substantively meaningful controls.
The specific categories of tabulation have been dictated by sample size and population distribution. The two racial categories displayed are white (88 percent of the sample) and blacks and others (12 percent). A separate set of tables details years of schooling completed, using the categories of less than high school (about 20 percent of the sample), high school graduate to 14 years (about 52 percent of the sample), and 15 years or more (about 28 percent). At older ages, the sample of highly educated persons is very thin, particularly for women. This has made the more conventional cutoff of a college degree impossible to implement. Race. Because the two components of worklife estimates, mortality and labor force behavior, are known to vary by race, the estimates themselves must also do so if appropriately tabulated. The new tables based on 1979-80 data now allow us to quantify the lifetime relationship between race and labor force involvement. As might be expected, the impact is striking, particularly for men.
Consider first the probabilities of moving into and out of the labor force. Among all men ages 16 to 64 who are outside the job market, whites are more likely to enter than are their minority counterparts. (See table 5.) Among those already in the labor force, blacks and others are the more likely to withdraw. The pool of inactive minority members incidence of labor force mobility at all ages.
The result is that minority men are estimated to average 4.3 labor force entries and 3.9 withdrawals per lifetime, while white men average 3.9 entries and 3.6 withdrawals. (See table 4.) Based on the observations for the reference period, the worklife expectancy of blacks and others was nearly 7 years shorter than that of whites (32.9 years vs. 39.8 years). Minority men spent an average of just 50 percent of their lives in labor force activity, compared with 56 percent for whites. This difference was all the more striking because whites tended to live longer, allowing them greater potential for both a longer worklife and post-retirement leisure. Far more blacks and others were likely to die before retirement (31.7 percent as against 26.7 percent for whites). Stated differently, although minority men could expect to spend fewer years in the labor force, their additional periods of inactivity were more likely to occur during prime working ages.
The racial differentials in worklife expectancy were less distinct for women. At most ages, it was minority rather than white women who were the more likely to enter the job market, if inactive. (See table 5.) However, they were also the more likely to withdraw from economic activity.
One apparent difference by race involved the childrearing years. Neither black nor white women showed strong tendencies to withdraw from the job market to have children. However, the data pointed toward a "fertility troubh," although weak, in the labor force attachment of white women. Contradicting the patterns observed at other ages, white women in their 30's showed a stronger prpensity to leave the labor force than did their minority counterparts, and those 35 to 44 showed a stronger tendency to reenter. Although the timing of midlife labor force withdrawal differed by race, estimates of lifetime entries and exits for the two groups are surprisingly similar. (See table 4.) On balance, white women averaged 2.3 more years of worklife (29.7 years vs. 27.4 years), but this is largely a reflection of their greater longevity.
Education. The new tables reveal a clear and direct relationship between years of schooling and duration of labor force involvement. As noted earlier, the size of the differential is probably understated. There has been no attempt to estiamte the impact of eucation on health and survival.
The mechanism whereby education affects worklife duration is probably occupational selection. Although the link between schooling and occupation is imperfect, many occupations are closed to persons who have not met minimum educational requirements. Therefore, breaking the population into three educational strata effectively breaks it into clusters of occupations for which certain levels of training may be necessary.
The new tables reveal a decided employment "payoff" for time spent in school. During the prime working ages, men with 15 or more years of schooling are roughly half as likely to leave the job market, if active, as are those without high school diplomas. (See table 4.) If inactive, their probability of labor force entry is approximately twice that of the least educated group. Over a lifetime, the most educated class of men averages slightly more entries and exits than do those without high school diplomas, but most of this turnover occurs relatively early, while many individuals are still in school. After age 25, thse men can anticipate fewer transitions in either direction.
Over a lifetime, the average man with 15 years of schooling or more can expect to work 6.5 years longer than his classmate who left high school before graduation (41.1 vs. 34.6). The same increment to education will have twice as much impact on the worklife duration of a woman, adding an average of 12.6 years to her economically active life (34.9 vs. 22.3 years).
Table 6 isolates the impact of education during three periods of the worklife cycle: the early and middle phases and the preretirment years. It displays the number of years the average person can be expected to work during each such phase, by sex and years of schooling completed.
At younger ages, education has a two-pronged effect on men: While failure to earn a high school diploma costs the individual about a year and a half of worklife between the ages of 20 and 39, remaining in school also imposes a cost in terms of forgone employment opportunities. However, among the group ages 40 to 59, the payoff from education is very evident. Those completing 15 years of school or more can expect to work 1 year longer than high school graduates, and 3 years longer than those who did not graduate. Even though higher education, with its greater compensation returns, may ease the financial strain of retirement, it seems to engender a sense of "career commitment" in many men which holds them in the labor force. (This is evident in the separation rates in table 5.) Examples of this phenomenon include self-employed career professionals such as attorneys and physicians, who are reputed to remain active long after most wage and salary workers have retired. An additional effect of schooling seems to be that--among those who have "retired," at least in terms of their principal job--the most educated are the most likely to return to work in some capacity, as reflected in accession rates. Finally, if educational attainment is positively correlated with good health and longevity, untreated health problems may discourage economic activity among the least educated, least affluent groups, further widening the worklife gap associated with schooling. Thus, in the final phase of the work cycle, the most educated group remain active 1.6 years longer than high school graduates and 3 years longer than those who never finished high school.
The work patterns of women vary more widely than those of men. Consequently, education has a stronger potential impact on female worklife behavior than on that of males. The new tables show this effect to be the greatest during the prime working ages. Between the ages of 20 and 39, women face fundamental tradeoffs among schooling, childrearing increase with job skills. During this phase of life, the woman with 15 years of schooling or more is likely to work nearly a year longer than the high school graduate, and 4.5 years longer than her classmate who left high school early. The differential remains, and in fact widens, throughout midlife. During the next 20 years of her life cycle, the highly educated woman is likely to work 2.9 years longer than the high school graduate, and 5.9 years longer than the nongraduate. The tables suggestt that in relationship between education and retirement patterns is looser for women than for men. As with men, the most educated show the least inclination to retire early. (See table 5.) However, once they have done so, these women are less likely than men with comparable training to reverse their decision. (as evidence, compare accession rates of the most educated men and women in table 5.)
This latest worklife study, based on a larger sample of individuals than had been used previously, has enabled us to examine two new dimensions of worklife behavior. It has also provided more complete measures of movement into and out of the labor force than were previously possible.
During the period between 1977 and 1979-80, the worklife expectancy of adult men held relatively steady, while that of women continued to edge upward. For both sexes, there were indications that many retirement decisions were being reversed. However, because the model does not measure hours of labor force involvement, the workyears remaining to older persons may in fact be less "intense" now than they were at the beginning of the decade.
Race seems to have more bearing on the worklife patterns of men than of women. The tables confirmed that minority men are both more likely to leave the labor force and less likely to reenter than are whites. The racial differential for women affects timing of movement more than it does overall volume.
The more important factor affecting worklife patterns of women is educational attainment. Using the ecategories displayed here, we find that women appear to reap twice as much "payoff" from additional schooling as do men. Their additional training appears to drive up the opportunity costs of alternative activities, encourging longer and more continuous careers for those who have pursued higher education.
Opportunity costs also appear to play an important role in the retirement process. For both sexes, higher education is associated with later retirement. Among the men who do retire, the most educated are most prone to reenter the work force. The swifter, more permanent retirement pattern of persons without high school diplomas may be due, in part, to health differentials by educational attainment, mentioned but not fully controlled for in this study.
|Printer friendly Cite/link Email Feedback|
|Author:||Smith, Shirley J.|
|Publication:||Monthly Labor Review|
|Date:||Aug 1, 1985|
|Previous Article:||Recent trends in unemployment and the labor force, 10 countries.|
|Next Article:||The FCMS contribution to nonlabor dispute resolution.|