The Length of Working Life

Transcription

1 The Length of Working Life BY SEYMOUR L. WOLFBEIN The past ten years have witnessed major advances in the application of demographic techniques to the study of the labour-force problems in the United States. The severe depression of the nineteen thirties focused attention on the trends in the size and composition of the working population; the war years underscored the importance of these trends, especially for studies of the labour supply available to meet the requirements of the war effort; and the post-war period has re-emphasized this general area of work in assessing the problems posed by a return to a peacetime economy. The study of labour-force trends provides an excellent illustration of the underlying importance of demographic factors to many other specific disciplines. The size of the labour force at any given time, for example, depends in large part on the aggregate size of the population, its age-sex structure and on other demographic characteristics. Labour-force trends, in addition, depend on such specific variables as births, marriages, mortality and migration. Recent changes in the labour force of the United States have, in fact, been strongly influenced by demographic trends. For example, the current pronounced shortage in certain occupations, such as teachers and nurses, is traceable in no small measure to the extremely high marriage and birth rates of the post-war years. This paper will deal with one specific aspect of labour-force inquirythe measurement of working life. It illustrates not only the relevance of demographic factors, but adapts the technique of the life table to a measurement of the work-life span. The concept of 'working life' is, of course, not a novel one. Since the eighteenth century at least, actuaries, economists, and sociologistsin attempting to arrive at some measure of the 'capitalized' value of human lifehave made explicit or implicit estimates of the duration of productive life. One of the most recent of these appears in the work of Dublin and Lotka, The Money Value of a Man. With the rapid growth of the aged population in recent decades, increasing attention has been given to the economic problems of dependency in old age. Estimates of length of working life, as compared to the total life span, are pertinent to this problem. In this context, for example, Woytinsky in 1938 published estimates of the 'expected period of work' for gainful workers.^ More recently, John Durand has developed estimates of 'average number of years in the labour force' in his study of The Labor Force in the United States, , published last year, In addition to the interest in the average work-life span, those concerned with problems of labour-market dynamics also have need for related measures of the rates of separation from the labour force at different ages. The Bureau of Labor Statistics, for example, in attempting to assess long-range employment opportunities ^ W. S. Woytinsky, Labor in the United States, Social Science Research Council, [286]

2 THE LENGTH OF WORKING LIFE 287 in various occupations, found that in many skilled and professional fields, the greatest proportion of job openings results from the need to replace losses due to death or retirement. However, despite the importance of this factor, very little direct information on the rates of separation from different occupations has been available. In the absence of more specific data, patterns of labour-force separations over age have been developed for men in certain broad population groups, on the basis of mortality data and the Census labour-force statistics. Once developed, it became apparent that this body of data could be significantly integrated in the framework of the life table. Structure of the tables of working life A working-life table for total males in 1940 is shown in Table i. In common with the conventional life table, the table of working life starts with an initial group of 100,000 at birth, and follows it through life, subject to a pattern of attrition determined by a specified set of mortality rates. Since the focus here is on the span of working life, the life table starts with age 14, the age at which measurement of the labour force begins under present Census definitions. At age 14, for example, the stationary population (L^) has already been reduced from 100,000 to 92,115. Beginning with age 14, the life-table population is also exposed to two other factors: the probability of accession and of separation from the labour force at different ages. Unlike the mortality data, however, no ' vital statistics' are available as to the number of different persons entering the labour force or separating, each year, by age. In fact, within the framewqrk of the current United States labour-force statisticswhich define labour-force membership in terms of activity in a given week, rather than in terms of statusno precise measurement of labour force 'births' or ' deaths' is feasible. Some individuals can, of course, bound their workinglife span by certain definite events, such as the time of high-school graduation or the age of retirement on a pension. Most workers, however, pass through transitional phases both at the early and later stages of working life, when their attachment to the labour force is marginal and intermittent. The high-school student, working occasionally after school hours or during the summer vacation, and the aged pensioner, who still performs odd jobs, are illustrations of this sort of marginal situation. For purposes of life-table construction, net changes in labour-force activity over age, at a given seasonal level served as the basis for estimating entries and withdrawals. Age-specific rates of labour-force participation (i.e. worker rates) were developed after considerable adjustment and smoothing, from the Census data for April Net rates of accession and separation from the labour force were then derived from the differences in successive worker rates. What was assumed, for example, was that, given a static pattern of worker rates, a cohort of workers aged 60 in 1940 would acquire the lower worker rates of persons aged 61 as they became a year older. From the difference in successive worker rates in combination with the mortality data, it was possible to derive the probabilities of separation due either to death or retirementusing the latter term to signify all separations for causes other than death. ; With this basic assumption, the development of the table of working life represents

3 288 SEYMOUR L. WOLFBEIN a fairly straightforward application of standard life-table techniques. A number of departures from conventional usage were introduced, however, partly because of the nature of the data and partly in view of the end uses of the tables. These will be noted in the following description. Table i, for total males, is based on the 1940 pattern of labour-force participation. It will be noted that some of the functions normally included in the standard life table have been omitted, i.e. the mortality rate (qx), the number living at beginning of year of age (Z^.), the number dying (d^), and the cumulative man-years of life remaining (T^). These and corresponding labour force functions were omitted from the life table in the interest of compactness. In most instances they can readily be derived from the given data. In column 2 is shown the' stationary population', the number living within a given year of age, of 100,000 born alive (L^.). These values were derived from the Census Bureau's United States Abridged Life Tables, The 1939 tables were used in preference to the more recent United States Life Tables for because they were the only official United States life tables presenting mortality rates by urbanrural residence. The mortality rates shown in the 1939 tables are somewhat higher than those shown for , at most ages. However, for ages 14 and over, the maximum effect on the life expectancy function {e^ is a difference of o-i6 years at age 14, with considerably smaller differences at subsequent ages. Columns 3 and 4 show the number (LW^) and percentage {W^) of the life-tahle population in the labour force within each year of age. These columns were based on single-year worker rates derived from the published Census labour force statistics ^ for 1940, after adjustment and smoothing. As shovwi in Table i the worker rates for ^, total males rise sharply in the latter teens and early twenties when most young men!s enter the labour force. They continue to rise slowly, however, to a peak at about -^ age 32. Between the mid-thirties and mid-fifties the proportion of men in the labour n force drops very gradually, but then declines rapidly due to the cumulative effect of :;^ retirements. i Column 5 (A^) shows the net accessions to the life-table labour force between ilt successive years of age as a ratio to the life-table population (LJ in the base year. ^ It is computed from the net increase between successive worker rates, up to the v; maximum at age 32, after accounting for losses to the labour force due to death > during the year..1 _ r M/ _ m Vr _ nd\ ra i Columns 6-8 show the probabilities of separations from the labour force, due to all causes (Q%) and due separately to death (Q^) and to retirement (Qx)- The probability of separations {Q^) is defined as the net separations from the life-table labour force between successive years of age, of those in the stationary labour force in the base year. It differs conceptually from the mortality rate (q^) which relates the number dying between two exact age intervals to the population at the beginning of the interval. The use of the stationary labour force as a base was designed, in the present tables, to facilitate application of the derived separation rates to available population and labour data which are all expressed in terms of attained age. For

4 (I) Year of age X to * ' ^ _ 97-^ (a) (3) (4) No. living of 100,000 bom alive I r\ population 92,115 91,968 91,812 91,638 91, ,008 90,771 90, , ,883 88,581 88, ,619 87,269 86,902 86,520 86, ,777 84,268 83,729 83,160 82,553 81,901 81,205 80,458 79,661 78,809 77,895 76,921 75,883 73,616 72,379 71, ,261 66, , ,099 56,129 54, ,804 40, , ,855 20,391 18, , ,605 5, , (In In labour Force No. LWx year of age) ,192 21, ,240 65, ,690 82,646 83,824 84, ,828 84,789 84,643 84,409 84,132 83,812 83,452 83,060 82,636 82,173 81,664 81,109 80, ,162 78, ,865 75,996 75, , , ,144 66,733 65, ,902 60,057 58, S , , , , ,058 19,217 16, ,266 10, , , a I I'ercentage of opulation P Wx Table i. Total males I O 1-6!! 0-8 O-5 (5) the labour 1000 in opulation) 1000 Ax ' (6) (7) (8) Separations from the labour force (per 1000 in labour force) Due to all causes 1000 Ox Due to death 1000 Qf (Between years of age) I Hi III Due to retirement loooqj ^ Q S-8 A. mm II (9) (10) Average no. of remaining years of Life I ex ^abour force )articipation ««, (At beginning of year of age) J Il-O 11-1 xo S O' c.q ft Q li yb * I I Q'O 20' i _ ^ V Q.I Io IS' no a.f '- ~ H 3-6 i "-* i'3 2' I 2'0 1 'Q e

5 MH.:. 290 SEYMOUR L. WOLFBEIN ages 32 and over (i.e. after the maximum worker rate has been attained), the annual probability of labour-force separation for persons in the labour force in a given year of age was computed as a ratio of the difference between the stationary labour force in successive years to the labour force in the base year For ages it was assumed that net separations were due entirely to death; and therefore these values were derived from the product of LWj.{Qj.), where Q^. was defined as the mortality rate between successive years of age ^-^". (3) In deriving the probabilities of separation due to death or retirement after age 32, a further assumption was made. In the absence of differential mortality rates for workers and non-workers, it was necessary to assume that the age-specific death rates for men in the labour force approximated those for the population as a whole. It is likely that this assumption tends to overstate the labour-force death rate, since those outside the labour force include a large percentage of sick and disabled persons. The differential is likely to be greatest at earlier ages, when, however, only a small proportion of men are outside the labour force, and probably relatively small at the upper ages when many factors other than disability influence labour force status. Given the population mortality rate (Q^) and the separation rate, Q^., it was then possible to derive the separate probabilities of death and of retirement. Solving algebraically, the respective formulae for the probabilities of death and retirement were: _ Q^(2-Q%) - (5) In the last two columns of the table are shown the two life expectancy functions: the average number of years of life remaining (e^) and the average number of years remaining in the labour force (iw^). The total life expectancy function is identical with that shown in the conventional life tables, and is computed by dividing the cumulative man-years of life in the given year and all succeeding years by the number living at the exact year of age 0) The average number of years remaining in the labour force for ages 32 and over is similarly derived by dividing the cumulative man-years in the labour force in the given year and all succeeding years, by the number in the labour force at the beginning of the year of age: 0) I (7) S^

6 THE LENGTH OF WORKING LIFE 291 The population and labour force at the exact age interval was estimated by linear interpolation from the corresponding L^ and LW^ values. This assumes an even distribution of deaths and retirements within each year of age. In the case of mortality, the assumption of even distribution within each year of age for the age span under consideration is consistent with prevailing actuarial practice, though it involves a slight statistical bias. In the case of retirements, the bias is probably somewhat greater, particularly at pivotal ages such as 65, but further refinements did not seem warranted in view of the lack of basic data. Since the working life expectancy has been defined in these tables as the average number of years of working life remaining to a group of persons in the labour force at a given age, a modification of the above formula was needed for ages 14-31, inclusive, to eliminate the effect of entries into the labour force in the following years. For this purpose, a set of hypothetical values of the stationary labour force (LW^) and of the labour force at the beginning of each year of age (Iwj), were computed by applying the maximum worker rate, of age 32, to the life-table populations for ages LW^; = L,(W^3,); K = Uw,,). (8) Thus, for purposes of computing the life expectancy function at age 14, it was assumed that all entries into the labour force had already occurred by age 14, and that the labour force at age 32 was smaller than that at age 14 by the number of deaths in the intervening years. Working-life patterns in the United States The working table permits a number of interesting comparisons of the length and pattern of working life, at different times and among various population groups. Highly significant is the contrast between total life expectancy and working-life expectancy, as an indication of the problem of old-age dependency. Under 1940 conditions, the average male worker aged 20 could expect to live for an additional 46-8 years and to remain in the labour force for an additional 41-1 years. He would therefore have to provide for an average of almost 6 years of life in retirement. If he survived and was still in the labour force at age 60, his average life expectancy would extend until age 75, but he could still count on spending about 6 years outside the labour force. These, of course, are averages: a large proportion of men die before they have an opportunity to retire, while others may spend extended periods in retirement. Over past decades, there has been a steady widening in the gap between total-life and working-life expectancies. There has been a tendency for gains in work-life expectancy to lag behind the progress in extending the biological life span. The shift from a rural to an urban economy and related occupational trends, combined with the extension of old-age pension programmes, have had the effect of lowering the age at which men drop out of the labour force. The net result of these changes is indicated by a comparison of the 1940 life tables with estimates of working-life expectancy for white males in 1900.^ In 1900, the ' For 1900, the mortality rates used were those of the original death registration states, while the worker rates were for all white males in the United States, adjusted for comparability to current Census definitions.

7 SEYMOUR L. WOLFBEIN 20-year-old white male worker had, on the average, an additional life span of 42-2 years, or about 5 years less than in His working-life expectancy of 39*4 years was, however, only i J years less than in On the average, therefore, he could expect slightly under 3 years outside the labour force, as compared with 6 years in Between 1940 and 1947, this trend towards an extension of the span of dependency in old age was halted, at least temporarily. The average life expectancy increased by a full year over this period. At the same time, the wartime boom and post-war prosperity had the effect of permitting increased labour-force participation of older men, particularly in the age group As a result, the average future working-life span of a young man aged 20, under 1947 conditions, also increased by a full year. These comparisons suggest alternative patterns for future trends in the work-life span. A resumption of pre-war trends on the one hand, would mean a rapid widening of the period of dependency and would add correspondingly to the costs of old-age benefit programmes. On the other hand, if worker rates of older men are maintained at 1947 levels, there will be very little change in the proportion of the adult life span spent outside the labour force. The comparisons by colour and residence developed in the working-life tables for 1940 also present a number of significant contrasts which may be mentioned briefly here.^ Men in rural areas begin working at an earlier age than urban workers, and once in the labour forcehave a longer average working life. Thus, the white rural worker at age 20, could (under 1940 conditions) expect to continue working for an additional 43-2 years, on the average, or about 2^ years more than his urban counterpart. This is due both to the greater average longevity of the adult rural population and to the lower rate of retirement. The family farm offers a much more flexible outlet for continued gainful activity even at advanced ages than does urban industry, and as a result a much larger proportion of men in rural areas continue in the labour force until death or serious disability. Even more striking are the contrasts between whites and non-whites. The nonwhite youth, at age 20, has an average future work-life expectancy about 5I years less than the white worker. Higher mortality rates for non-whites between the twenties and fifties largely account for this sharp difference. This differential narrows markedly over the period of middle age and reverses in favour of the non-whites in the late fifties. The predominantly rural composition of the non-white labour force at these later ages, with its low retirement rates and apparently favourable mortality, largely accounts for this shift. Alternative measures of working-life expectancy The method used in estimating the future working-life span, in the present paper, is only one of several possible alternatives. It is particularly important, in the developmental phase of research in the aspects of labour force analysis, to note the ^ A Bulletin, showing Tables of Working Life for 1940 and 1947 classified by colour and place of residence, and containing a detailed exposition of methodology, as well as an analysis of the substantive findings, is available from the U.S. Bureau of Labor Statistics, Washington 25, D.C

8 THE LENGTH OF WORKING LIFE 293 alternatives (or supplementary methods) and to obtain a consensus of opinion, if possible, as to their desirability and feasibility. (1) Work-life expectancy of the population. In the present table, the average future years in the labour force {iw^^ were computed for men in the labour force. This definition seemed most consistent with the life-table concept. However, it is equally possible to compute the future working-life span of all persons living at a given year of age, regardless of their labour-force status at that time. Essentially this approach was used by Durand in estimating the 'average number of years in the labour force * for men of different ages.^ Using this method, for example, it would be possible to estimate the future working-life potential of a newly born infanta measure of considerable intrinsic interest. This definition, moreover, takes account of trends in the age of entry into the labour force, as well as of separation, and therefore permits a more comprehensive measure of trends in the total work-life span. However, this measure seems somewhat less meaningful at the upper ages, when a large proportion of the population has already left the labour force. (2) Alternative periods of labour-force activity. For certain purposes, measures of entries to and separations from the labour force based on concepts other than the current Census labour-force definitions might be desirable. As noted previously, at both age extremes of the population a large proportion of persons reported in the labour force are engaged in casual and part-time jobs. Measures based on major activity during the survey period (i.e. student, retired, worker) would yield a more realistic pattern of labour-force 'births' and retirements. The use of 'full-time* equivalents, by age, would also have a similar effect. Whereas the preceding methods would tend to reduce the estimated work-life span, a maximum measure would be obtained by including in the labour force all persons participating in the labour force at any time during the year, rather than at a given seasonal level of activity. In view of the seasonal fluctuations in labour-force participation and the large volume of shifting in worker status among teen-aged youths and, to a lesser extent, among older workers, the effect of this definition would be to produce an earlier average entry age and a later average age of retirement. (3) Estimates of work-life span based on gainful employment. For purposes of measuring income-earning potential, estimates of working-life expectancy can be constructed based solely on periods of gainful employment for pay or profit, rather than on labour-force participation. Excluded, on this basis, would be periods of unemployment or of employment as unpaid family workers. (4) Generation life tables. Finally, working-life tables could be developed based on a historical cohort analysis, rather than on a static pattern of mortality and labourforce participation. If long-range projections of worker rates are available, tables could be developed using available mortality projections to show the prospective working-life span of persons currently entering the labour force. The development of such tables would be handicapped, however, by the absence of comparable labour-force data in sufficient age detail in Census years prior to 1940, and by inherent difficulties in long-range projections of trends in labour-force participation. * John Durand, The Labor Force in the United States, , pp J D III iii 20

9 294 SEYMOUR L. WOLFBEIN Further research on patterns of working life The brief review of the work done so far in this field serves to highlight the fact that it is still in an early developmental stage. From the point of departure of the overall patterns of working life for men, much more intensive research is indicated, in a number of directions. Following is a summary of the major lines of research which suggest themselves: (1) Working-life tables for women. The much more complex pattern of entries and separations of women from the labour force over age is affected by such demographic variables as marriage, childbirth, widowhood or divorce, and by the diminution of household responsibilities as children age and leave the household. Tables measuring the effects of these demographic factors on the changing labour-force status of women over age are currently being developed by the Bureau of Labour Statistics. (2) Occupational attrition tables. The development of age-specific rates of separation from different occupations is an essential link in any realistic forecasting of replacement needs and potential job openings. The use of overall rates of death and retirement, from the tables of working life, represents only a crude approximation of the probable rates of separation for these causes from any given occupation. Basic statistics must be developed on occupational mortality and on the differential rates of retirement, by at least broad occupational groups. However, in order to estimate gross occupational replacement needs we need to know much more, too, about mobility (i.e. the rates of shifting among occupations, by age): a vital area of labour force research which is still largely unexplored. (3) International comparison of working life. The use of the working-life tables and related measures will permit ready geographical comparison of the working potential of populations. To the extent that adequate labour-force data are available, such comparisons may be developed between different countries, as well as for regions, states or local areas within a nation. In addition to the other variables involved, allowance should be made in these tables for the pattern of migration, where this is a significant factor. (4) Analysis of causes of labour-force attrition. In the present work-life tables, separations from the labour force, other than death, have been subsumed under the general heading of 'retirement'using this term in a very broad sense. Included among the causes of retirement, as thus defined, are disability, entry into an institution (for reasons other than disability), voluntary retirement on a pension or annuity, or lack of employment opportunities. With the growth in our aged population and of the problem of old-age dependency, intensive research will be required as to the relative importance of these and other factors. These basic data bear the same vital relationship to the analysis of trends in working life as the statistics of 'causes of death' do to the trends of mortality.

10