Click for next page ( 172


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 171
I`~ Work Experience, I Job Segregabon, and Wages MARY CORCORAN, GREG J. DUNCAN, and MICHAEL PONZA Women are a vital part of today's labor force, and work is clearly an important part of their lives. Women constituted more than two-fifths of the labor force in 1978, and almost 60 percent of all women aged 18 to 64 were employed in 1978. Almost all women work at some point in their lives, and their earnings are often necessary to ensure ad- equate family support. In 1978 nearly two- thirds of the women working were either presently unmarried or married to men earning less than $10,000 per year (in 1977. There is considerable evidence that men's and women's work participation patterns aider with men working continuously after completing school and women moving in and out of the labor force to accommodate family and child-rearing duties. Women earn considerably less than men do. Since 1930 the median salary of full-time, full-year women workers has been about 60 percent of the median salary of men who work full This paper was supported by a grant from the Alfred P. Sloan Foundation. ~ These figures are taken from U.S. Department of Labor, Women's Bureau, 10 Facts on Women Workers, Washington, D.C., August 1979. 171 time, full year. Women and men also have very different occupations. Treiman and Hartmann (1981) show that 70 percent of the men and 54 percent of the women in the labor force are concentrated in occu- pations dominated by their own sex. Unlike men, women are heavily concentrated in a few job categoriessecretarial work, sales, teaching, nursing, and various service oc- cupations. The most prominent economic explana- tion linking labor supply patterns and wages is human capital theory.2 Human capital it- self is defined as worker skills or qualifica- tions acquired through schooling or on-the- job training. An individual worker's stock of human capital can be increased by the proc- ess of investment. Investments have an op- portunity cost (in terms of forgone earnings as well as of the direct costs of training) and a return (in the form of higher subsequent earnings). Human capital theory particu- 2 Human capital theory is quite similar to the func- tionalist theory of Davis and Moore. It has been argued that this theory underlies much of the empirical work in social stratification (see Horan, 1978~.

OCR for page 171
172 MARY CORCORAN, GREG [. DUNCAN, AND MICHAEL PONZA larly emphasizes investment in formal schooling (Becker, 1975) and in on-thejob training (Mincer, 1974~. Workers are pre- sumed to choose freely among jobs with dif- ferent amounts of training, and wages are presumed to reward past investments in ed- ucation and training in a similar way for all workers. In recent years human capital theory has been expanded to deal with the structure of female wages. Some of its proponents have argued that the sex division of labor within the home generates sex differences in pat- terns of investment in work-related human capital and that this in turn generates sex differences in wages and the sex segregation of occupations (Mincer and Polachek, 1974; Polachek, 1976, 1979, 1981; Mincer and Ofek, 19821. These arguments have focused par- ticularly on sex differences in patterns of labor force participation. This paper investigates human capital theory's predictions about the relationships between patterns of work, wages, and job segregation. The paper is in four major parts. The first summarizes human capital theo- retical models. In the next section we use 13 years of data from the Pane} StudY of Income Dynamics (PSID) to describe and compare men's and women's patterns of work participation. These comparisons focus on aspects of lifetime work experience that are hypothesized to be important for women duration of work and nonwork periods, the extent to which labor market experience in- volves part-time work, and the sex typing of experience (i. e., the extent to which past experience was in female-dominated occu- pations). The third section reviews past re- search on the extent to which different as- pects of work experience influence wages and sex typing of a person's current lob. ant} it also includes the results from our own analyses of these issues based on 13 years of PSID data. Finally, we discuss the the- oretical and policy conclusions drawn from our reviews of past research and from our own research. WORK HISTORY, WAGES, AND JOB SEGREGATION: THEORETICAL MODELS Work Experience and Earnings In the human capital model, investments in on-thejob training are considered to be critical determinants of wages (see Becker, 1975; Ben-Porath, 1967; Mincer, 1974; and Rosen, 1972~. On-thejob training has a cost, since time spent in training is assumed to be time diverted from production, and pro- duction presumably determines earnings. On-thejob training also has a return in the form of higher later earnings. The following function describes this hypothetical rela- tionship: ~~ F:. = E, + 2, rC' = Ye + C`, (~1) i=o where En is earnings capacity in year t, Es is earnings that would be received in the absence of any postschoo! training, Ci is the dollar cost of investments in human capital in the ith year, Y' is earnings in the tth year, C' is dollar cost of investments in the tth year, and r is rate of return to investments in human capital. If we assume that total benefits of an in- vestment increase as the payoff period in- creases and that the marginal costs of in- vestments are upwardly sloping in a single time period, it can be shown that a profile of investment ratios (CilE`) that are large at first and then decline over time maximizes the present value of expected lifetime earn- ings (see Ben-Porath, 19671. That is, the proportion of one's earnings capacity in- vestec3 in on-thejob training will be high in the early years and then will decline rapidly. The human capital mode! assumes that workers freely choose among a variety of jobs each with a different combination of training and productive work. It generally views training and productive work as mu- tually exclusive activities, and, thus, ac- cording to the mociel, employers will pay

OCR for page 171
WORK EXPERIENCE, ,~ SEGREGATION AND WAGES . _ 173 less for jobs with training than for similar jobs that do not provide training. An im- plication of the mode} is that wages grow with experience because workers are ac- quiring additional subs as they increase their experience and so their wages grow not just because of seniority. Labor Force Withdrawals and Wages Mincer and Polachek (1974) extend the human capital mode! to account for the pos- sible depreciation of human capital that may result from the discontinuity of women's work experience. They argue that during periods of labor force withdrawal for child-bearing and child-rearing, prolonged nonparticipa- tion in the paid labor market can cause the skills acquired at school and work to become less valuable. The following function adjusts the basic human capital wage mode} to account for depreciation or obsolescence effects: ~~ E`= Es + ~ (rC'BE`), i=l where E`, Es, r, Ci are defined as in Eq. (1), bi is the depreciation rate of human capi- tal in year i, and Ei is earnings capacity in year i. The total benefits of investments in on- thejob training increase with the length of the payoffperiod but decline with the length of periods of nonparticipation that follow in- vestments. This suggests that optimal in- vestment patterns will differ depending on the continuity of market activities. Contin- uously employed workers should concen- trate investments early in their careers. Workers who interrupt their work careers will defer investments in on-thejob training until they reenter the labor market after completing these activities so as to minimize the loss from depreciation. Since such work- ers have a shorter payoff period, their over- all volume of investment should be lower than that of workers who remain continu- ously in the labor force. Mincer and Ofek (1982) have since re- vised this initial mode! to account for "res- toration" or "repair" of depreciated human capital. They argue that the "reconstruction of (previously eroded) occupational skills is more efficient than the construction of new human capital." That is, it costs less to repair human capital than to build it. This resto- ration phenomenon leads Mincer and Ofek to distinguish the short-run and long-run consequences of nonparticipation. In the short run (say, the first year following an in- terruption), one would expect sharply lower wages than those received just prior to the interruption, followed by a period of rapid wage growth during which human capital is restored. Thus, the long-run effects on wages of nonwork time may be considerably smaller than the short-run effects. Since the empirical work of Mincer and Ofek and our own replication of it show that wage "rebound" following an interruption is an important phenomenon, it is useful to consider alternative explanations of it. Cor- (2) coran (1979) and Corcoran and Duncan (1979) suggest that time out may lead to a tem- porary reduction in wages because of tem- porary mismatches between worker skills and jobs. Women workers lack complete infor- mation about job opportunities when they do return to the labor force, and it takes time for them to discover jobs that are best matched to their skills. Employers also have imperfect information about the productiv- ity of their new employees, and the learning process for them is time-consuming.3 One 3 Morgensen (1978), Jovanovic (1979), and Prescott and Visscher (1980), for example, explain that earnings rise with experience with a firm because firms learn about worker productivities in various jobs (instead of workers' acquiring skills through experience). This learning process results in the more senior workers being matched more accurately to jobs commensurate with their skills than less experienced workers. Better job matches allow the senior workers to exhibit higher productivity on average, and if the market rewards productivity, these differences may account for their higher average earnings.

OCR for page 171
174 MARY CORCORAN, GREG J. DUNCAN, AND MICHAEL PONZA common mechanism for this sorting process is to hire new workers in at low wages but then to promote them rapidly as they suc- cessfully complete their probationary pe- riods. In neither ofthese cases are the work- ers "restoring" depreciated skills in the Mincer-Ofek sense. Wage increases accom- pany improved information of employees about their job opportunities or improved information of employers about the produc- tivity of their employees. Note that Mincer and Ofek have altered the original Mincer and Polachek mode! substantially. Since depreciated human cap- ital can be restored, it no longer follows that intermittent workers will necessarily defer investments in on-thejob training until all interruptions are over. This decision will de- pend on the relative sizes of the cleprecia- tion and restoration effects.4 Similarly, the relative sizes of these two effects will also determine the Tong-run wage costs of labor force withdrawals. If these long-run costs are small, then depreciation may account for little of the wage gap between men and women. Part-Time Work Experience and Wages Women are considerably more likely than men are to work in part-time jobs, a fact that may lead to considerable differences in the amount of on-thejob training women ac- quire and, therefore, in their relative wage growth. The most general human capital theories (Heekman, 1976; Blinder and Weiss, 4 A qualification suggested to us by Jacob Mincer is necessary here: For the intermittent worker, each in- terruption carries with it a positive probability of not returning to the labor market. Thus, the expected pay- off period is diminished by more than the interruption each time it looms. So although wage loss due to de- preciation can be made up, the intermittent worker's decision about whether and when to invest will depend jointly on the relative sizes of the depreciation and restoration effects and on the probability of returning to work. 1976) do not make unambiguous predictions about the effect of part-time work on human capital investment and wages, but there are reasons for believing that less training is ac- quired in part-time work than in full-time work. First, because part-time work means fewer hours in the labor market than full- time work does, women who expect to work part time in the future have a shorter ex- pected work life and hence less incentive to invest in on-thejob training. In this case both the overall volume of investment and the rate of investment would be lower for those who plan to work part time than for those who plan to work full time. If current part-time work patterns are associated with the likelihood offuture part-time work, then current part-time workers will be making fewer investments. Second, if employers suspect that part-time workers are more likely to leave than are full-time workers, they might restrict training opportunities in part- time work. Employers would be most likely to restrict opportunities for firm-specific training. Finally, just as it is argued that skills depreciate during periods of nonwork, skills could depreciate more (or appreciate less) during part-time work than during full- time work, since part-time work involves fewer hours of work (i.e., more hours of nonwork). The depreciation from nonuse would be greater if the nature of part-time work precluded workers from maintaining their market skills. If formal training is scheduled when part-time workers are not at work, then there will be less wage growth resulting from the acquisition of new skills for them. Two sources of data with crude direct measures of on-thejob training do show a positive relationship between work hours and training. Duncan and Hoffman (1979) found with the 1976 wave of the PSID that adult workers aged 18 to 64 who worked less than 20 hours per week reported training periods attached to their jobs that were only about half as long as those of workers working be- tween 40 and 50 hours. Stafford and Duncan

OCR for page 171
WORK EXPERIENCE, [OB SEGREGATION, AND WAGES 175 (1979) found qualitatively similar, although less statistically significant, differences in training by labor supply for workers in the 1975-1976 Time Use Study. Labor Force Withdrawals and Job Segregation: Human Capital Explanations The 1974 Economic Report of the Presi- dent speculated that sex differences in pat- terns of work participation may be the cause of the sex segregation of jobs. This line of reasoning has been extensively developed by ZelIner (1975) and Polachek (1976, 1979, 19811.5 Since Polachek's explanation sub- sumes Zeliner's model, we will concentrate on his mode! in the following discussion. Polachek (1981) defines atrophy as the loss of earnings potential that occurs when skills are not continuously used. He shows that if the cost of labor force withdrawals (the atro- phy rate) varies across occupations, and if lifetime labor force participation differs across individuals, then a worker will choose "that occupation which imposes the smallest pen- alty, given his desired lifetime participa- tion." This model treats the "lifetime as a unit."6 Thus, this mode! implicitly assumes that workers tend to work in the same sort of occupation throughout their lives or at least over long periods of time. Lifetime work participation is assumed to be exogenously determined, and atrophy rates are assumed to vary across occupations. This mode! provides a human capital ex- planation for the sex segregation of the labor market. If work skills do atrophy during s England (1981, 1982) provides an extensive discus- sion of these models. This section is informed by her work. 6 Polachek notes that "this assumption can be relaxed by posing the problem within a dynamic control frame- work" but goes on to say that "even within such a framework the same conclusions hold for occupations chosen at a given stage of the life cycle" (Polachek, 1981, p. 64~. Note that this relaxation still implies oc- cupational immobility over a life-cycle stage. withdrawals from the labor force, then it is rational for women who expect to take time out from the labor force to work in fields where there is less chance of atrophy i. e., in fields with low depreciation rates but also with low returns to experience. Thus, such women wid experience less atrophy than will women who expect more continuous work participation. By selecting jobs that are easy to leave and reenter, women can thus more easily combine the clual demands of career and family. Polachek's mode} can explain sex segregation only if typically "female" jobs are those where there is the least atrophy. Note that if depreciates! skills can be re- stored (as Mincer and Ofek argue), this weakens the force of Polachek's arguments. Webster's defines atrophy as "a wasting away or progressive decline." Thus, the cas- ual reader might assume that Polachek's atrophy rate is equivalent to Mincer and Polachek's depreciation rate. But atrophy, as defined by Polachek, picks up two things depreciation (i.e., reduction in work skills due to nonuse) and forgone appreciation (i.e., the loss in expected earnings growth due to missing a year of work).7 Depreciation and the growth of earnings with experience are quite different proc- esses. Depreciation implies that the level of work skills is lower following an interruption 7 We are grateful to Siv GustaEson for first pointing out this distinction to us in a personal conversation in 1978. England (1981, 1982) is the first author who clearly makes this distinction in a published paper. England first referred to the loss in expected earnings growth as "forgone appreciation," and provides an excellent discussion of Polachek's models. Zellner's explanation of the sex segregation of oc- cupations rests solely on forgone appreciation. England (1981) points out: "Zellner assumes that occupations can be divided into those that offer high initial salaries and flat earnings profiles and into those with low initial salaries and steep earnings profiles. Women, because of their shorter expected work lives, will be more likely to maximize lifetime earnings in the occupations with high initial salaries and flat wage growth i.e., in 'fe- male' occupations."

OCR for page 171
176 MARY CORCORAN, GREG [. DUNCAN, AND MICHAEL PONZA than it was just prior to that interruption. If this earnings loss is long lasting, it is obvious why women who expect prolonged labor force withdrawals should enter fields with low de- preciation rates. Polachek's mocle! also im- plies that women who expect prolonged withdrawals should enter fields with high initial salaries but fairly flat earnings growth rates. This second decision only makes sense if we assume that, all else equal, jobs with high earnings growth pay less initially than jobs without such earnings growth. (Eng- land [1981] makes this point quite clearly.) To summarize, Polachek provides an in- genious human capital explanation for job segregation. If Polachek's general mode} is correct, then women who anticipate pro- longed nonwork time shouicl work in fields with low atrophy rates and should experi- ence less depreciation and less wage growth than do otherwise similar workers. Pola- chek's model also has several implications for the nature of typically female and male occupations. First, since women's choice of a "female" or"male" occupation reflects life- time participation plans, we would expect that the sex typing of women's occupations change little over a prolonged period of time. In addition, we should find that deprecia- tion and/or earnings growth will be lower in "female" occupations and that women who expect discontinuous careers will choose "female" rather than "male" occupations be- cause discontinuity is penalized less. Be- cause of this choice, women with discontin- uous work careers will be concentrated in "female" occupations. Occupational immo- bility and low wage growth are also pre- dicted by any job segregation model that presumes that women are locked into a set of female-dominated jobs that do not pro- vide productivity-enhancing experience. WOMEN'S WORK AND OCCUPATIONAL HISTORIES The human capital models summarized above predict that women's low wages result from a low overall volume of work, inter- mittent work participation, and part-time work. Most job segregation models, human capital or otherwise, implicitly assume con- siderable immobility between "female" and "male" jobs. As a first step toward testing these models, we use 13 years of data from the PSID to assess the accuracy of these descriptions of women's work behavior and occupational immobility. Patterns of Labor Supply, 1967 to 1979 Every year PSID respondents report their own and their spouse's work hours. Cor- coran et al. (in press) examined labor supply for adult men and women, aged 23 to 47 in the first year of the panel, who lived in their own households.8 As expected, there were dramatic differences between the sexes in the frequency and regularity of work and in the extent of part-time or part-year work. Differences between the races were much less dramatic within the groups of men and women. Black women acquired more ex- perience than white women did, while black men acquired less of it than did white men. Between 70 and 80 percent of the two groups of women were absentirom the labor force for at least 1 of the 13 years.9 The ~ In terms of the PSID sample, this group consists of all individuals who were household heads or wives in each of the 13 years. Eliminated from this analysis are children and the small group of other relatives of the household head (e. g., brother or sister). The lower age restriction was imposed to avoid sample selection problems associated with the decision to leave the pa- rental home and form one's own household. The upper age restriction eliminated from the sample individuals who would have reached the early retirement age of 62 by the end of the panel period. The results are reported in Corcoran et al. (in press). 9 We used 250 hours during a calendar year to define whether an individual was in or out of the labor force during that year, and we use 1,500 hours to separate part- from full-time workers. This procedure has its disadvantages. An individual with a 40-hour-per-week job who drops out of the labor force altogether for six months out of a calendar year will be classified as a part-time worker during that year without having a spell of nonwork. In one sense this individual was a

OCR for page 171
J WORK EXPERIENCE, IOB SEGREGATION, AND WAGES 177 comparable fractions for white and black men were about 10 and 16 percent, respectively. Even when they did work, women were much less likely to work full time. Less than one-tenth of these adult women worked full- time during the entire 13-year period. The total volume of work experience ac- quired by men was much higher than for women. On average, men worked in almost twice as many of the years as women did, in nearly twice as many weeks as women did, and for nearly three times as many hours as women did during this period. The part-time nature of the work of women was highlighted when we examined hours worked per week and weeks worked per year during the work spells. The average work week of men exceeded 40 hours, amounting to 46 hours for white men and 43 hours for black men. In contrast, white and black women averaged 36 hours per week during work spells. Similar differences showed up in the number of weeks worked per year, with men averaging 47 weeks and women averaging 42 weeks. As a result, the tote] number of hours averaged by men dur- ing their work spells was almost twice as high as for women during their spells. Patterns of Occupational Segregation, 1975 to 1979 Both Polachek's (1981) model and seg- mented labor market models assume little mobility between "male" and "female" oc- cupations over a prolonged period of time. Yet England (1982) reports that the corre- lation between percent female in detailed census occupation coding of first job and percent female in detailed coding of 1967 job is only .39 for women aged 30 to 44 years in 1967. This suggests there may be consid- erable mobility between "male?' and "fe- male" job sectors. We further tested this assumption of in- tersectoral immobility by calculating pat- terns of occupational segregation over the years 1975 to 1979. Our measure of occu- pational segregation is based on 2-digit oc- cupation and 2-digit industry categories. Thus, it has the advantage of accounting for both occupational and industrial segmen- tation by sex.~ We define a female-domi- nated job as an industry-occupation group with more than 50 percent women workers. Industry-occupation groups with less than 50 percent female workers are designated male-dominated jobs. To investigate the dynamics of job seg- regation over the period from 1975 to 1979, we selected a sample of women aged 23 to 57 in 1975 who worked in the first and last years of that period and who may or may not have worked during the three years in between. ii About 70 percent of white women workers held female-dominated jobs in 1975. If job segregation was completely rigid, then we would expect to observe that same frac- tion spending all of their working years in jobs dominated by women. Table 10-1 shows that this is clearly not the case. Only half of the white women spent all of their working years in the five years between 1975 and 1979 in jobs dominated by women. Less than one-sixth of these white women spent all of their working years in jobs dominated by men, leaving more than one-third who switched job types at least once. Switches between female- and male-dominated jobs for black women were almost as common full-time worker and in another sense this individual experienced a spell of nonwork during that year. Our A Our procedure for determining whether a given measure considered part-time workers to be those either job was "female-dominated" or "male-dominated" is working a limited number of hours per week or those detailed in Corcoran et al. (in press). working during only part of the year. Our measure of ii As before, the sample consists of household heads nonwork spells required that such spells be long enough ' -' ' ~' ' ' '' to take an individual away from work for virtually an entire year. and wives in this age range and thus excludes a small number of adults who are related to the head of the household in some other way.

OCR for page 171
178 MARY CORCORAN, GREG I. DUNCAN, AND MICHAEL PONZA TABLE 10-1 Dynamics of Occupational Segregation Subgroup Women Who Worked at Least 2S0 Hours in 1975 and 1979 Occupational Change Fraction spending all working years in female-dominated jobsa Fraction spending all working years in male-dominated jobs Men All White White Black Black .07 .18 (1,563) (606) All .08 (2,169) Fraction switching at least once in either direction Fraction of those in female-dominated jobs initially who switched to male- dominated jobs .51 (871) .15 (871) 34 . (871) .31 (647) .61 (538) .09 (538) .31 (538) .25 (423) .52 (1,409) .14 (1,409) 34 . (1,409) .30 (1,070) .79 (1,563) .15 (1,563) .60 (207) .67 (606) .18 (606) .37 (106) .78 (2,169) .16 (2,169) .56 (313) Fraction of those in male-dominated jobs initially who switched to female- .44 .55 .45 .09 .13 .09 dominated jobs (224) (115) (339) (1,300) (456) (1,756) NOTES: Table reads: 51 percent of the 871 white women who worked at least 250 hours in 1975 and 1979 spent all of their working time in female-dominated jobs. The number of observations is given in parentheses below each estimate. a A job is designated as female dominated if the percentage of women comprising it is greater than or equal to 50. Otherwise it is designated as male dominated. SOURCE: Panel Study of Income Dynamics. 31 percent of the black women were codes! as switching from one type to the other. These figures on the extent of switching between job types mix together both kinds of changes. A more relevant statistic on the issue of whether women who take female- dominated jobs remain in them is the frac- tion of women who began in female-clomi- natec! jobs and switched out of them. That fraction is 31 percent for white women and 25 percent for black women. These figures on switches between job types deserve careful scrutiny. On the one hand, they are likely to understate the true amount of movement between job types be- cause the time span over which such changes can be observed is limited to only five years. i2 i2 Women who worked continuously in the same sec- tor between 1975 and 1979 but switched in 1980 or had switched in 1974 are classified here as persistent residents in one sector. Also, while most of the women in this sample worked in every one of the five years, In addition, our classification procedure for identifying female- and male-dominated oc- cupations is a crude one and undoubtedly misses some true switches that would be caught with a more refined set of occupa- tional and industrial codes. On the other hand, errors in the coding of occupation and industry may create the appearance of a switch when in fact there was nones It is impossible to say whether the net effect of these considerations is to increase or to de- crease the estimated extent of switching be- some did not work in some of the middle three years, giving them fewer than five years in which a switch can be observed. i3 Appendix Table B.l in Corcoran et al. (in press) sheds some light on this by showing comparable mo- bility figures for the case when male-dominated jobs are defined as less than 40 percent female, and female- dominated jobs are defined as greater than 60 percent female. Hess mobility is found with this more restric- tive definition, but the extent of mobility is still sub- stantial.

OCR for page 171
WORK EXPERIENCE , . _ SEGREGaTION. AND WAGES 179 tween job sectors. However, it is almost cer- tainly true that the extent of switching is substantial, a fact that is inconsistent with Polachek's assumption of occupational im- mobility and with any other labor market mode} based on rigid segmentation by sex. This result also suggests that analysts should be wary of using a woman's current occu- pation as a measure of her past occupational history. WORK HISTORY AND WAGES: EMPIRICAL EVIDENCE The Depreciation Effect Empirical evidence on whether deprecia- tion lowers wages consists mostly of cross- sectional data where earnings of different individuals with different work histories are compared, after statistical adjustment to make the individuals as similar as possible. This evidence has produced contradictory results about the size of the depreciation effect- i.e., about the extent to which wages de- cline with time out of the labor force once one controls experience and tenure. Mincer and Polachek (1974) reported Mat 1967 wages dropped by 1.2 percent per year out of work for white married women aged 30 to 44 with children. Sandell and Shapiro (1978) repli- cated the Mincer-Polachek analysis after correcting for coding errors in women's re- ports of employment behavior. They re- ported that wages cleclined only 0.4 percent per year of nonparticipation and that this effect was insignificant. Corcoran (1979) rep- licated the Mincer-Polachek analysis for a national sample of wives aged 30 to 44 with children, taken from the 1975 PSID and ob- tained similar results to those of Mincer ancl Polachek. But Corcoran (1979) and Cor- coran and Duncan (1979) also reported that the decline in wages was much smaller (0.6 percent per year out of work) for working women in a broader age range (18 to 64 years). These results suggest that wages of married women aged 30 to 44 are more af- fected by labor force withdrawals than are wages of women in a broader age range. A recent paper by Mincer and Ofek (1982) suggests that some of these inconsistencies in past research arise because cross-sec- tional analyses tend to confound the short- run and long-run effects of nonparticipation. Mincer and Ofek use eight years of National Longitudinal Survey (NLS) data to explore how time out of the labor force affected wage growth for a sample of women aged 30 to 44 in 1967 who were married sometime dur- ing 1967 to 1974. They found a large short- run loss in wages immediately following an interruption (ranging from 3.6 to 8.9 per- cent per year out of work), followed by a period of rapid wage growth. Estimates of long-run wage losses were moderae (0.4 to 1.1 percent per year out of work). A repli- cation and extension ofthis analysis was con- ducted by Corcoran et al. (in press) using 13 years of information from the PSID. Short- run depreciation effects were estimated to range from 2.5 to 4.7 percent per year de- pending on the exact form of the mode! and on the definition of the sample when the age range was identical to the one used by Mincer and Ofek (30 to 44~? and the effects were estimated to be similar in magnitude when the age range was extended to be- tween 23 and 47 years. Long-run deprecia- tion was estimated to be between 1.0 and 1.5 percent per year in the replication, with some of these coefficients not statistically significant at conventional levels. Ibis recent work reconciles the disparate estimates of depreciation. Ike past analyses of wage effects of withdrawals based on cross- sectional data are likely to pick up both short- run and long-run effects. Married working women aged 30 to 44 who have interrupted work are likely to have recently returned to the labor force, and so short-run effects may have a large weight in analyses run on this group (e.g., that of Mincer and Polachek, 19741. Analyses run on women in a broader age range (e.g., that of Corcoran, 1979; and that of Corcoran and Duncan, 1979) are likely to put more weight on the long-run effect

OCR for page 171
180 MARY CORCORAN, GREG r. DUNCAN, AND MICHAEL PONZA - and so will provide lower estimates of de- preciation. The Restoration Effect In their original paper, Mincer and Po- lachek (1974) suggested that the optimal timing of investment in on-thejob training would differ depending on the continuity of market activities. In particular, workers who interrupt their work careers for nonmarket activities will defer investments in on-the- job training until they reenter the labor mar- ket after completing these activities so as to minimize the loss from depreciation. Min- cer and Ofek (1982) have considerably re- vised this hypothesis by arguing that "de- preciated" or "erodecl" human capital can be cheaply and rapidly "restored" soon after labor market entry. They show that postin- terruptior~ wages grow at roughly 2.5 per- cent per year of experience, on average, and that growth rates in the first year following an interruption range from 5.8 to 6.4 per- cent per year depending on the exact spec- ification of their model. This growth rapidly erases estimated short-term losses from the depreciation associated with short spells out of the labor force and is much larger than the wage growth of comparable continuous workers. They further demonstrate that growth in tenure accounts for less than half of this wage growth and interpret this to mean that the remainder is due to growth (repair) of general training. The nature and causes of wage rebound following work interruptions are crucial ele- ments in understancling the wage conse- quences and job choice of female labor sup- ply patterns. If depreciation is quickly repaired, it no longer follows that intermit- tent workers will defer investments in on- thejob training until all interruptions are completed. Thus, the observation that the wages of women grow more slowly in the years following the completion of schooling because of the reduced incentives to invest in human capital may no longer hold. It also weakens the plausibility of the reasoning that intermittent workers will concentrate in fe- male jobs. The Mincer and Ofek (1982) estimates of restoration came from the estimation of a cross-sectional wage equation. Corcoran et al. (in press) were able to use more complete information about the amount and sex typ- ing of work experience before and after work interruptions and also estimated a wage- change equation a specification with sev- eral statistical properties that make it pre- ferred to a wage-ZeveZ equation. Since this work has only recently been completed and addresses many of the important issues con- sidered in this paper, we summarize our analysis here. Readers interested in the de- taiTs are referred to Corcoran et al. (in press). We selected the adult women in the PSID sample and used the 13 years of work history obtained for them. We developed our wage equation by identifying the first (F) and last (L) wage observation for all women who worked at least 2 of the 13 years. 14 A cross- sectional wage equation at time F would be of the form: ZnW`F = aOF + alFSiF + a2FeOiF + 0~3FhOiF + ~FXiF + IF + IF ~ (3) where W`F is the wage rate of the ith indi- vidual in time F. aOF is a constant in time F. SiF is the education level of the ith indi- vidual in time F. eOiF is years of work ex- perience for the ith individual in time F. hoiF is years of nonwork for the ith individual in time F. XiF is a vector of observed produc- tivity-related characteristics for the ith in- dividual in time F. ZiF iS a vector of unob- servable individual-specific productivity- related characteristics, workplace charac- teristics, and labor market differences for the ith individual in time F. and ifs iS the 14 Throughout this section we use 250 hours as the cuto~point to distinguish individuals in the labor force from those out of the labor force. The sensitivity of these results to change in this definition are detailed in Corcoran et al. (in press).

OCR for page 171
WORK EXPERIENCE, JOB SEGREGATION, AND WAGES 181 stochastic disturbance term for the ith in- dividual in time F. PSID information on the work history be- tween F and L can be used to distinguish: en, years of work experience accumulated between F ant! L since the last completed interruption; e*, years of work experience accumulated between F and L prior to the most recent completed interruption; hi' years of nonwork between F and L during most recent completed interruption; and h*, years of nonwork accumulated between F and L prior to most recent completed interrup- tion. The cross-sectional wage relationship at time L (allowing the parameters to change) is given by InWiL, = aloe + Apsis + (X2reoiF + a3~hOiF + a4e*i + (X5eLi + a6h*i + Aphid + ~UrXi~ + ~~ZiL. + Din (4) The short-run depreciation and rebounc! effects are given by parameters a7 and as, respectively. Subtracting Eq. (3) from Eq. (4), suppressing the subscript i, denoting changes from F to L as "A," and adding and subtracting Arose, OAF, ant! U~ZF results in the following general equation for wage change: AlnW= Mao + a~r/`s + AalSF + ~a2eOF + ~a3hOF + a4e* + Uses + a6h* + a7hL, + ,u~,AX + /\ I1XF + BUZZ + A~TrZF +~. (5) If one assumes that the cross-sectional ef- fects of the explanatory variables are invar- iant between F ant! L and, further, that the unmeasured characteristics remain constant for the same individual, then the wage change equation simplifies to: AInW= auras + a4e* + aSeL, + ash* + a7hl + ,u~LliX + AD (6) Although the dependent variable in Eq. (6) is wage change rather than wage level, the parameters on the experience variables (a4 - a7) correspond to the parameters in the cross-sectional Eq. (4~. The key advantage to the change formulation is that estimates of these parameters are free from the sta- tistical problems caused by retrospective re- ports and by unchanging, unmeasured var- iables correlated with the included (measured) explanatory variables. An addi- tional advantage is that many more women meet the requirements of working at least 2 of the 13 years than work in a single year, and, therefore, selection bias problems are much less severe in estimating change Eq. (6) than in estimating a cross-sectional equa- tion. 15 Table 10-2, columns 1 and 2, shows es- timates of wage-change Eq. (6), which is the longitudinal analogue to the Mincer-Ofek cross-sectional equation. The work segment following the most recent interruption cent was entered quadratically to allow for a more rapid growth at first. Results show the es- timated rate of wage growth immediately following the last interruption to be a little over 5 percent per year for white women and 8 percent for black women, with the rate of growth declining to zero after about 10 years for both groups, which is close to the maximum observer! value for the eL var- iable in the sample. Depreciation during the most recent interruption is estimated to be- tween 4 and 5 percent per year, so the initial wage rebound following an interruption more than makes up for the wages lost during a year out of the labor force. Effects of Prospective Interruptions on Wage Growth Since the profitability of investments is affected by the length of time over which is See Corcoran et al. (in press) for more detailed discussions of the development of the wage-change equation and of selection bias adjustments.

OCR for page 171
182 MARY CORCORAN, GREG I. DUNCAN, AND MICHAEL PONZA TABLE 10-2 Basic Wage Growth Regression Independent Variable White h*: years out of labor force prior to most recent interruption ha: years out of labor force during most recent interruption e*: years in labor force prior to most recent reentry Black 016 `.030y 046** t.016' .030** `.008' .080** `.026y .0041* `.0019y White Black -.015 `.~o' -.070** `.020y .023* `.011' .077* <.032' - .0045* (-arm) -.124 `.103y .osst `.032' - .006 `.008y .ooos .0090) .035 521 en: years in labor force during most recent spell e 2 L NT79: Did not work in 1979 NT79*h, NT79*el. NT79*e~,2 R2 (adjusted) Number of observations .016 `.029) 038** `.013' .012t (-~ .0s2* `.o~' .0027t .0016) .021 .057 837 521 .013 .029 035* .015 .012 (.oo9) .05 .027 0025 (.0015 017 (.o99) 030 .029 .043 (.056 008 .0056) .024 837 NOTE: Standard errors are in parentheses. * Significant at .05 level. ** Significant at .01 level. Significant at .10 level. SOURCE: Panel Study of Income Dynamics. benefits are received, the human capital mode} predicts that otherwise identical workers will invest less if they anticipate labor market withdrawals than if they do not. This assumption is also implicit in Po- lachek's argument that workers who antic- ipate time out will choose occupations with low atrophy rates. Sandell and Shapiro (1980) test this proposition by estimating whether NLS women who expected to be out of the labor force at age 35 had flatter experience- earnings profiles before then than did women who expected to be working at age 35. Al- though most of their key parameter esti- mates are in the expected direction, none are significant at the 5 percent level for white or black women. This hypothesis is an important one for the human capital model; it deserves testing in the context of the wage-change models developed here. Since we know which women in the PSID sample were not work- ing at the end of the 13-year period, we can test directly whether such workers'jobs pro- vided them lower wage growth and lower depreciation. In contrast to the self-re- ported intentions of respondents used in the articles listed above, this procedure tests for the effects of actual labor force behavior in period t + 1 on wage profiles in period t. We did this by creating a dummy variable (NT79) for whether did not work in 1979, interacting this dummy with he, en, and e2, and adding these four variables to the basic wage-change Eq. (61. The results of this analysis are reported in Table 10-3, col- umns 3 and 4. In general, white women who did not work in 1979 had the same wage increment for additional years of experience as women who did work in 1979 and similar wage loss with time out as did otherwise similar white women who were working in 1979. But one result for black women does conform with human capital predictions.

OCR for page 171
interruption h,: years out of labor force during most recent interruption Years of full-time e* Years of part-time e* Years of full-time e' (Years of full-time eJ-squared Years of part-time en .034 (.013) .016* (.008) .001 (.011) .029 (.008) .001 (.010) .037* (.015) .024* (.008) .018 (.011) .030 (.010) .017 (.013) -.035** (.013) .007 (.007) - .002 (.011) .062** (.021) - .0036* (.0018) - .021 (.026) (Years of part-time e-squared .0016 (.0025) Years of full-time en in male- dominated jobs Years of part-time e, in male- dominated jobs Years of full-time en in female- dominated jobs WORK EXPERIENCE, TOB SEGREGATION, AND WAGES 183 TABLE 10-3 Effects of Part-Time Work and Female-Dominated Work on Wage Growth Independent Variable White Black White Black White h*: years out of labor force .023 - .005 .022 - .026 .021 prior to most recent (.029) (.030) (.029) (.029) (.029) -.018 (.029) .049 (.015) .034 (.008) .023* (.011) .037 (.024) .008 (.0021 .150* (.029) .014 (.003) .036 (.013) .005 (.008) .003 (.012) .026* (.012) .048* (.015) .015 (.001) .008 (.011) .008 (.016) Years of part-time eL in female- - .004 - .003 dominated jobs - (.012) (.014) Number of cases 837 521 837 521 837 521 R2 (adjusted) .025 .049 .028 .096 .023 .044 - .009 (.019) .020* (.009) .024 (.028) .021* (.011) * Significant at .05 level. ** Significant at .01 level. SOURCE: Panel Study of Income Dynamics. Black women who were not working in 1979 exhibited no wage loss during prior labor force withdrawals. i6 Their past wage return to experience, however, did not differ from i6 The depreciation estimate for black women who did not work in 1979 is the sum of the coefficient on he term (-.070) and the coefficient on the NT79*h~ term (+.059). that of otherwise similar black women who were working in 1979. Part-Time Work Experience and Intermittency Corcoran (1979) and Corcoran and Dun- can (1979) included a retrospective measure of experience in a cross-sectional equation for a national sample of women. They report

OCR for page 171
- 84 MARY CORCORAN, GREG 1. DUNCAN, AND MICHAEL PONZA that part-time work experience was signifi- cantly less valuable than full-time work ex- perience. Jones and Long (1979) used data from the NLS to estimate two cross-sectional wage equations that included interactions be- tween work experience segments and whether the work was for part of a week. Although the signs of the coefficients they estimated were consistent with the hypoth- esis that part-time work leads to slower wage growth, only two of the 12 coefficients were statistically significant at conventional lev- els. Their measures of the part-time nature ofthe work segments were very rough, how- ever, and may have biased some of the coef- ficient estimates. Corcoran et al. (in press) also investigated the effects of part- and full-time work on the wage growth of women workers with some simple modifications to their basic wage- change equation. In Eq. (6), full- and part- time years in the e* and en segments were not distinguished. Since the volume of work hours was ascertained for each of the 13 years under investigation, that information can be used to classify years of experience that in- volved part-time work Jess than 1,500 hours) and full-time work (1,500 hours or more). Four variables were formed with this infor- mation: (1) the number of years of e* that were part-time (e*-part), (2) the number of years of e* that were full-time (e*-ful0, and (3) the number of years of en that were part- time (er-part), and (4) the number of vears of en that were full-time (er-ful0. This de- composition of e* and e~yields the following equation: A1,nW= a,L,/\S + a6h* + a7hL, + a~e*-part + age*-full + a~OeL,-part + a~eL-full + FLAX + AD. (7) As with the more basic measures of e* and en, all four of these new variables are ob- tained in each of the annual interviews and do not rely on retrospective reports by either women workers or their husbands. The results of the estimation of the aug- mented wage-growth equation are shown in Table 10-2, columns 1 to 4. Full-time work floes indeed appear to be associated with significant wage growth, while part-time work does not. When the two measures of en are entered linearly, the wage growth associ- ated with full-time experience is positive and significant for both white and black women, while the wage growth associated with years of part-time work in the most recent spell of employment was insignificant for both groups of workers. With years of experience prior to the most recent spell of nonwork (e*), the full-time work variables have larger coefficients than do the part-time variables, although these differences were not signif- icant at conventional levels. A parabolic specification for the en measure gives ex- pected results for white women there is a parabolic rebound for full-time but not part- time work. For black women, there is par- abolic wage rebound for part-time work- a puzzling result. Sex Differences in Work History and the Sex-Based Wage Gap Two sets of analysts have extensively ex- amined the relationship between work his- tory and the sex-based wage gap on a na- tionally representative sample of women. Mincer and Polachek (1974) estimated that sex differences in experience and time not working accounted for about 45 percent of the wage gap between employed married men and women aged 30 to 44 years in 1966. About half of this difference was due to the depreciation effect. i7 Corcoran and Duncan }7 Mincer and Polachek used data from two different sources to make this comparison. This led to some inconsistencies between male and female variables. The sample of men was taken from the 1966 Survey of Eco- nomic Opportunity (SEO). This survey does not meas- ure either work experience or tenure directly. Men

OCR for page 171
WORK EXPERIENCE [OB SEGREGATION AND WAGES . . (1979), using a broader age range and a more extensive list of work history measures, found that sex differences in work history ac- counted for between one-third and two-fif;chs of the wage gap between working white men and working women aged 18 to 64 in 1975. This occurred largely because women had acquires] less tenure and were more likely to have worked part time. The depreciation effect and intermittence clid little to explain the wage gap between women and white men. Work History and Occupational Segregation: Empirical Evidence Polachek's (1981) argument implies that women choose "female" jobs because such jobs penalize discontinuous labor force par- ticipation less than "male" jobs do. This ex- planation presumes that there is immobility between occupations; that depreciation and/ or wage growth are lower in "female" jobs than in "male" jobs and that women who expect discontinuous careers are concentra- ted in "female" jobs. We have already demon- strated that there is substantial mobility be- tween "male" and "female" job sectors over a five-year period a result that contradicts a basic assumption of Polachek's explanation. The empirical evidence presented by Po- lachek for this argument is indirect. Pola- chek (1981) has shown that the probability of currently working in a given occupation (keened by 1-digit census categories) is af- fected by years out of the labor force and that the size of this effect differs by occu- pation. He has also demonstrated that the relationship between wage growth and years out of the labor force (home time) differs across occupations (i.e., occupations have different atrophy rates). He has further shown were assumed to have no interruptions. These two re- sults are not inconsistent. In both cases, sex differences in work history explain a large but not major part of the sex-based wage gap. Sample differences likely ac- count for differences in the importance of depreciation. 185 that there is a negative correlation between the effect of years out of the labor force on the probability of working in an occupation and the atrophy rate in that occupation. Even if we ignore the issue of mobility across occupations, Polachek's evidence does not unambiguously support his hypothesis. Take Polachek's first finding that home time affects the probabilities of currently being in a particular occupation. This finding can only explain sex segregation of jobs if women with extensive home time were more likely to work in female-dominated occupations than were otherwise similar women without ex- tensive home time. Polachek used his es- timates of these effects to obtain a projected population-wicle occupational distribution for women 30 to 44 if they had worked con- tinuously since school completion. He re- ported that the proportion of women profes- sionals anal managers (currently male- dominated fields) would increase and that the proportion of women in household and service work (currently female-dominated fields) would decrease. On the other hand, his figures indicated an increase in the pro- portion of women employed in clerical work (a female-dominated field), a decrease in women employed in crafts (a male-domi- nated field), and a decrease in the propor- tion of women in sales (an integrated field). i8 i8 Probably the best way to evaluate the quantitative importance of this evidence is to estimate the extent to which occupational sex segregation would be re- duced if women's home time were zero i.e., if men and women had the same work participation patterns. We estimated this by applying Duncan and Duncan's (1955) segregation index to Polachek's sample. This in- dex measures "the minimum proportion of one group that would have to be shifted for its occupational dis- tribution to be equal to that of the other." The seg- regation index for Polachek's sample is .50. Then we calculated this segregation index on the projected oc- cupational distribution calculated by Polachek under the assumption that women worked continuously. If his theory is correct, occupational sex segregation should be considerably reduced. Under the assumption that women do not withdraw from the labor force, the seg- regation index is .48 a reduction of only 2 percent.

OCR for page 171
186 MARY CORCORAN, GREG I. DUNCAN, AND MICHAEL PONZA England (1982) investigated the effect of home time on sex typing of current occu- pation more directly. She reports that the sex composition of most recent occupation and the sex composition of first occupation were uncorrelated with the proportion of total time employed since school completion for white women aged 36 to 50 years in 1973. i9 This result does not suggest a strong link between labor force discontinuity and the sex typing of current or first job. Now we turn to Polachek's second find- ing that the detrimental wage effects of home time vary across occupations. He cal- culated this by regressing the difference be- tween 1972 and 1967 wages on home time and other variables expected to affect wages. Unlike most economic studies of wage dif- ferentials based on the human capital model, Polachek examines dollar changes in wages rather than percentage changes. If all oc- cupations had identical percentage wage de- creases per year out of the labor force, Po- lachek would likely obtain differences in dollar wage change, with highly paid oc- cupations showing greater decline. Indeed, in Polachek's analysis, dollar wage changes are most negative for professionals, crafts- people, and managers the three highest- paid occupations. In order to study wage change between 1967 and 1972, Polachek must restrict anal- ysis to women who reported a wage in 1967 and in 1972. Thus, Polachek's sample is cho- sen on the basis of work behavior. This could possibly lead to selection bias problems when estimating effects of work behavior on wage change.20 A further problem is that Polachek's es- timate of effects of home time will be dom- inated by short-run effects, since he restricts analysis to a five-year period and only ex- amines the effects of withdrawals during that period. If lost skills were rapidly restored (as Mincer's and Ofek's t1982] and our re- sults suggest), then Polachek's estimates will considerably exaggerate the lifetime costs of time out. Finally, note that Polachek's atro- phy estimates pick up both depreciation and the forgone appreciation effects of fewer years of experience. Even if Polachek's evidence that occu- pations differ in atrophy rates were correct, this would only explain sex segregation if there were less depreciation of skills and lower returns to experience in femaTe-dom- inated occupations. England (1981, 1982) tested this assumption using both the NLS sample of mature women and a sample of women in a wider age range from the PSID. England's analyses have the two advantages of Polachek's analysis. She looks directly at the relationship between home time and sex typing of current job, and she examines de- preciation and returns to experience sepa- rately. England regressed the natural log- arithm of wages on experience, education, time out, and percent female in current oc- cupation (coded at 3-digit census level), and tested for significant interactions between experience and percent female and between time out and percent female. She reports that neither the depreciation rate nor re- turns to experience were affected by per- cent female in current occupation. This is fairly strong evidence against Polachek's ex- planation. Both Polachek's and England's empirical tests of the Polachek argument presume considerable immobility between occupa- ~9 England (1981) reports that Wolfe and Rosenfeld (1978) present some evidence that suggests a weak link between home time and sex composition of occupation. 20 This sample selection bias is a general problem for analysts of women's wages. At any point in time only about half of all adult women are in the labor force. Restricting the sample to women who worked in two specific years as Polachek does eliminates even more women from the sample. The wage-growth analysis of Corcoran et al. eliminates only about one-fifth of the sample, since it requires only that women work at least 2 of the 13 years under investigation. When we rep- licated our wage-change model for women who worked in 1975 and 1979, the results were inconsistent with results from the larger and less restricted sample.

OCR for page 171
WORK EXPERIENCE, [OB SEGREGATION, AND WAGES 187 tions over time. Whether their use of cur- rent occupation as a measure of occupational history is appropriate depends upon the va- lidity of this presumption. As we have dem- onstrated, there is extensive mobility be- tween "male" and "female" job categories. This in itself is inconsistent with Polachek's model. But it also suggests that the use of current occupation as a proxy for occupa- tional history is inappropriate and may pro- vide misleading information about whether job choice is conditioned by expectations about future work or whether experience garnered in "female" jobs results in lower wage growth and less depreciation than does ~ . ~~ ,, . experience garnered in ma he Jo as. We used the longitudinal nature of the PSID to develop more direct tests of the following two predictions of the human cap- ital model: 1. Wage growth and depreciation are lower for work experience gathered in "fe- male'? jobs than in "male" jobs. 2. Women workers with extensive time out and frequent interruptions are more likely to have concentrated their work experience ~~ r i, in tema he Jo as. We find virtually no support for the predic- tions. To test the first proposition, we modified our basic wage-growth equation to include the sex typing of experience. In Eq. (7) we did not distinguish whether years in the en segment involved work in male-dominated or female-dominated jobs. Since both in- dustry and occupation were reported for 9 of the 13 years under investigation, we could classify years in er that involved work in "female" and "male" jobs.2i We combined this with the information on work hours to create four new variables: 1. the number of years of en, that were full-time in male-dominated jobs (e~-fuit-m~, 2. the numbers of years of er that were part-time in male-dominated jobs (e~-part- 3. the number of years of en that were full-time in female-dominated jobs (e~-fuil- fct), and 4. the numbers of years of el that were part-time in female-dominated jobs (e~-part- fd). This decomposition of eL yields the following equation: bind= o~~l~/\s + a~Oe*-part + o`~e*-fuii + ar,4eL,-full-md + aL~5eL,-part-md + 16eL,-fUi[-fti + cx~7e~-part-fcl. (8) As with the basic measures of en, these four new variables are obtained in each of the annual interviews and do not rely on ret- rospective reports by either women workers or their husbands. These variables also pro- vide more complete measures of the extent to which work experience was acquired in "female" jobs than do measures of occupa- tion that are taken at a single point in time. The results of estimating Eq. (8), shown in Table 10-3, columns 5 and 6, provide little support for the argument that wage growth is much higher for male-dominated work than for female-dominated work, especially for white women. A much more important fac- tor was whether the work performed in a 2} Industry is coded into 2-digit categories for 1971 to 1979. Occupation is coded into 1-digit census cat- egories for the years 1971 to 1974 and into 2-digit cat- egories for all the years thereafter. For each occupa- tion-industry subgroup, we calculated a measure of percent female. If there were more than 50 percent women in that subgroup, we called it a "female-dom- inated" job. (See Corcoran et al., in press, for a more complete description of this procedure.) Since we only have measures of occupation and industry for the last 9 years of the study, we do not break experience in e. (which tended to occur early on in 1967 to 1979) into "male" and "female" components.

OCR for page 171
188 MARY CORCORAN, GREG I. DUNCAN, AND MICHAEL PONZA particular kind of job was part time or full time. For white women a year of full-time, male-dominated work was associated with a 2.6 percent increase in hourly wages, while a year of full-time, female-dominated work was associated with a 1.9 percent increase in hourly wages. The differences between these two coefficients were not significant. For black women a year of female-domi- nated, full-time work was associated with a 2.1 percent increase in wages. This com- pares to a 0.8 percent increase for a year of fi~-time, male-dominated work. Again, the difference was not significant. Part-time work experience, whether in "male" or "female" jobs, had no significant effects on wages for either blacks or whites. We examined whether the sex typing of women's work experience affected the rate of depreciation during labor force withdraw- als by interacting the two labor force with- drawal measures (h* and had with a measure of the average percent female in each wom- an's occupation-industry combination over the 13-year period. These interaction terms were always insignificant when added to the wage-change Eq. (81. This suggests that de- preciation does not differ for "male" and `d ~ ,, . tema e Jo as. The 13 years of PSID data allow a direct test of the second proposition that work- ers who expect discontinuous labor force ca- reers will concentrate in "female" jobs. If this hypothesis is correct, then the sex typ- ing of work experience over the years 1967 to 1979 ought to be positively related to time out of the labor force in 1967 to 1979, in- termittency of work participation in 1967 to 1979 (measured by number of labor force withdrawals), and whether working in 1979. Results of this exercise (see Corcoran et al., in press) confirm England's finding of no relationship between discontinuity of work 22 The .024~3 coefficient estimated for years of part- time, male-dominated work for black women appears to be out of line with the other results. This estimate is based on a small number of observations, as reflected in its large standard error. and sex typing of concurrent occupation. None of these three measures of labor force discontinuity over 1967 to 1979 had a sig- nif~cant, positive relation to the sex typing of work experience over that period. SUMMARY AND CONCLUSIONS The wage change models of Mincer and Ofek (1982) and those from our own work yield similar results. Women who drop out of the labor force have lower real wages when they return to work than they had when they left work. However, the period following the return is characterized by rapid wage growth, and the net loss in wages from drop- ping out is small. This result reconciles the apparently contradictory findings from cross- sectional studies about the size of the de- preciation effect, because cross-sectional analyses pick up both short-run and long- run depreciation effects. Short-run effects are likely given more weight in an analysis of women aged 30 to 44 (a group likely to have recently completed labor force with- drawals) than in analyses run on women in a broader age range. How does this empirical evidence affect our understanding of the process of female wage determination? The observed wage loss and rebound pattern is certainly consistent with the Mincer-Ofek story of human capital depreciation and restoration. This pattern is consistent with other stories as well- the job mismatch argument and the probation- ary period argument. We do not have the necessary data to disentangle these argu- ments. Regardless of the reason, the rapid rebound of wage losses after labor force withdrawals means that the wage losses as- sociated with these withdrawals cannot ex- plain much of the male/female wage gap.23 23 The fact that women work fewer years and more part-time years than men work does explain a substan- tial (one-third to two-fifths) part of the wage gap be- tween men and women workers in a broad age range (Corcoran and Duncan, 1979~. Note, however, that the bulk of the sex-based wage gap differences is still unex- plained by male/female work history differences.

OCR for page 171
WORK EXPERIENCE, TOB SEGREGATION, AND WAGES 189 Women are often urged to choose part- time work rather than to stop work alto- gether to keep their "hands in." Our results provide little evidence that the wage con- sequences of these two alternatives differ. Part-time work experience is not rewarded for women particularly for white women. And the long-run wage penalties due to la- bor force withdrawals are small. The deci- sion whether to work full time or part time is considerably more important than is the choice between part-time work or no work. These part-time results, like the wage loss and rebound results, are consistent with several quite different labor market scena- rios a human capital mode! of lower train- ing during part-time work, an imperfect in- formation model, or an institutional model. Again, we do not have the necessary data to disentangle the competing explanations, since they each involve a different interpre- tation of the same employer behavior. We also investigated the human capital models that explain job segregation as the result of women's discontinuous work his- tory patterns. Such models emphasize two costs of discontinuous work participation: depreciation and forgone wage growth. The rapid restoration of wage losses in the period immediately following labor force withdraw- als suggests that the first cost might be quite small. For these human capital explanations to hold, three things must occur: (1) there should be considerable immobility between "male" and "female" job sectors, (2) wage growth and depreciation should be lower for work in "female" jobs than for work in "male" jobs, and (3) women with discontinuous work careers will be more likely to choose "fe- male" jobs than will women with continuous work careers. We find little evidence for any of these propositions. First, there is considerable mobility between "male" and "female" job types. We did not find that either wage growth or depreciation varied significantly with the sex typing of work experience. These results are consistent with England's cross- sectional work. Finally, women with dis- continuous work careers were no more likely to have worked at "female" jobs than were women with more continuous work expe- rience. These results also have implications for models of job segregation other than the human capital model. Many models of job segregation either implicitly or explicitly as- sume that there is rigid segmentation be- tween "male" and "female" job sectors and that there are fewer promotion and/or train- ing opportunities in the "female" job sector than in the "male" job sector. The analyses reviewed in this paper suggest these as- sumptions are likely wrong. REFERENCES AND BIBLIOGRAPHY Appelbaum, Eileen 1981 Back to Work: Determinants of Women's Suc- cessful Re-entry. Boston, Mass.: Auburn House Publishing Co. Arrow, K. 1972 "Models of Job Discrimination." In Racial Dis- crimination in Economic Life, edited by A. H. Pascal, pp. 83-102. Lexington, Mass: D.C. Heath. 1976 "Economic Dimensions of Occupational Seg- regation: Comment I." Signs, 1, 3, Part 2 (Spring):233-237. Beck, E. M., P. M. Horan, and C. M. Tolbert 1978 "Stratification in a Dual Economy: A Sectoral Model of Earnings Determination." American Sociological Review 43 (October):704-720. 1980a "Industrial Segmentation and Labor Market Discrimination." Social Problems 28 (Decem- ber):ll3-130. 1980b "Reply to Hauser, 'Social Stratification in In- dustrial Society: Further Evidence for a Struc- tural Alternative."' American Sociological Re- view 45 (August):712-718. Becker, G.S. 1957 The Economics of Discrimination. Chicago: University of Chicago Press. Human Capital: A Theoretical and Empirical Analysis, With Special Reference to Educa- tion. Second edition. New York: National Bu- reau of Economic Research. Beller, A.H. In "Occupational Segregation by Sex: Determi- press nants and Changes." Journal of Human Re- sources. 1975 Ben-Porath, Y. 1967 "The Production of Human Capital and the Life-Cycle of Earnings." Journal of Political Economy 75 (August):352-365.

OCR for page 171
190 MARY CORCORAN, GREG 1. DUNCAN, AND MICHAEL PONZA Bergmann, B.R. 1974 "Occupational Segregation, Wages, and Profits When Employers Discnminate by Race or Sex." Eastern Economic Journal 1 (April-July):103- 110. Blau, F.D. 1977 Equal Pay in the Office. Lexington, Mass: D.C. Heath. Blau, F.D., and C.L. Jusenius 1976 "Economists' Approaches to Sex Segregation in the Labor Market: An Appraisal." Signs: Journal of Women in Culture and Society 1, 3, Part 2 (Spring):181-199. Blinder, A., and Y. Weiss 1976 "Human Capital and Labor Supply: A Synthe- sis." Journal of Political Economy 84~3~:449- 472. Cain, G. 1976 "The Challenge of Segmented Labor Market Theories to Orthodox Theory: A Survey. " Jour- nal of Economic Literature 14, 4 (Decem- ber):l251-1257. Corcoran, M. 1978 "The Structure of Female Wages." The Amer- ican Economic Review (papers and proceed- ings) 68, 2 (May):165-170. 1979 "Work Experience, Labor Force Withdrawals, and Women's Earnings: Empirical Results Us- ing the 1976 Panel Study of Income Dynam- ics." In Women in the Labor Market, edited by C.B. Lloyd, E. Andrews, and C.L. Gilroy. New York: Columbia University Press. Corcoran, M., and G.~. Duncan 1979 "Work History, Labor Force Attachment, and Earnings Differences Between the Races and Sexes." Journal of Human Resources 14, 1 (Winter):3-20. Corcoran, M., G.J. Duncan, and M. Ponza In "Work Experience, Job Segregation, and press Wages." In Five Thousand American Fami- lies Patterns of Economic Progress, edited by G.J. Duncan and J.M. Morgan. Ann Arbor: Institute for Social Research. Doeringer, P.B., and M.J. Piore 1971 Internal Labor Markets and Manpower Anal- ysis. Lexington, Mass: D. C. Heath. Duncan, G.J., and S. Hoffman 1979 "On-the-Job Training and Earnings Differ- ences by Race and Sex." Review of Economics and Statistics 61, 4 (November):594-603. Duncan, O.D., and B. Duncan 1955 "A Methodological Analysis of Segregation In- dexes." American Sociological Review 20 (April):210-217. Edwards, R. 1979 Contested Terrain: The Transformation of the Workplace in America. New York: Basic Books. England, P. 1981 "Wage Appreciation and Depreciation: A Test of Neoclassical Economic Explanations of Oc- cupational Sex Segregation." Mimeo, Decem- ber. 1982 "The Failure of Human Capital Theory to Ex- plain Occupational Sex Segregation." Journal of Human Resources (Spring). Gronau, R. 1974 "Wage ComparisonsA Selectivity Bias." Journal of Political Economy 82, 6 (November/ December):1119-1143. Gwartney-Gibbs, P. 1981 Married Women's Work Experience: Intermit- tency and Sex-Typed Occupations. Ph.D. dis- sertation, University of Michigan. Hanushek, E.A., and J.M. Quigley 1978 "Life-Cycle Earnings Capacity and the OJT In- vestment Model." Discussion Paper 7904. Rochester, N.Y.: Public Policy Program, Uni- versity of Rochester. Hauser, R.M. 1980 "Comment on Beck et al., ASR, October 1978, 'Stratification in a Dual Economy."' Amer~can SociologicalReview 45(August):702-711. Heckman, J.J. 1974 "Shadow Prices, Market Wages and Labor Supply." Econometr~ca 42, 4 Ouly):679-694. 1976 "A Life Cycle Model of Earnings, Learning, and Consumption." Journal of Political Econ- omy 84~4~:Sll-S44. Hill, M.S. 1979 "The Wage Effects of Marital Status and Chil- dren." Journal of Human Resources 14, 4 (Fall):579-594. Horan, P.M. 1978 "Is Status Attainment Research Atheoretical?" Amerzcan Sociological Review 43(August):334- 341. Jones, E. B., and J. E. Long 1979 "Part-Week Work and Human Capital Invest- ment by Married Women."Journal of Human Resources 14(Fall):563-578. Jovanovic, B. 1979 "Job Matching and the Theory of Turnover." Journal of Political Economy 87, 5, Part I (Oc- tober):972-989. Jusenius, C.L. 1977 "The Influence of Work Experience, Skill Re- quirement, and Occupational Segregation on Women's Earnings." Jourrlal of Economics and Business 29, 2 (Winter):107-115. Kahn, L.M. 1980 "Wage Growth and Endogenous Experience." Industr?al Re1~tions l9, 1 (Winter):50-63. Landes, E.M. 1977 "Sex Differences in Wages and Employment: A Test of the Specific Capital Hypothesis." Economic Inquiry 15, 4 (October):523-538. Lazear, E. 1976 "Age, Experience, and Wage Growth." Amer- ican Economic Review (September):548-558.

OCR for page 171
WORK EXPERIENCE, TOB SEGREGATION, AND WAGES 19 Lloyd, C.B., and B.T. Niemi 1979 The Economics of Sex Differentials. New York: Columbia University Press. Medoff, J., and K. Abraham 1981 "Are Those Paid More Really More Produc- tive? The Case of Experience. " Journal of Hu- man Resources 16 (Spring). Mincer, J. 1974 Schooling Expenence and Earnings. New York: National Bureau of Economic Research. Mincer, J., and H. Ofek 1982 "Interrupted Work Careers." Journal of Hu- man Resources 16 (Winter). Mincer, J., and S. Polachek 1974 "Family Investments in Human Capital: Earn- ings of Women. " Journal of Political Economy 82, 2, Part 2 (March/April):S76-S108. 1978 "An Exchange: Theory of Human Capital and the Earnings of Women: Women's Earnings Reexamined." Journal of Human Resources 13, 1 (Winter):118-134. Morgensen, D. 1978 "Specific Capital and Labor Turnover." Bell Journal of Economics, 2 (Autumn):572-586. Oppenheimer, V.K. 1970 The Female Labor Force in the United States: Demographic and Economic Factors Govern- ing Its Growth and Changing Composition. Westport, Conn.: Greenwood Press. Piore, M.J. 1971 "The Dual Labor Market: Theory and Impli- cations." In Problems in Political Economy: An Urban Perspective, edited by D.M. Gordon, pp. 90-94. Lexington, Mass.: D.C. Heath. Polachek, S.W. 1976 "Occupational Segregation: An Alternative Hypothesis." Journal of Contemporary Busi- ness 5, 1:1-12. 1979 "Occupational Segregation Among Women: Theory, Evidence, and a Prognosis." Women in the Labor Market, edited by C.B. Lloyd, E.S. Andrews, and C.L. Gilroy. New York: Columbia University Press, pp. 137-157. . 1981 "Occupational Self-Selection: A Human Cap- ital Approach to Sex Differences in Occupa- tional Structure." Review of Economics and Statistics 63, 1 (February):60-69. Prescott, E., and M. Visscher 1980 "Organizational Capital." Journal of Political Economy, 88, 3 Oune):446~61. Rosen, S. 1972 "Learning and Experience in the Labor Mar- ket." Journal of Human Resources VII (Sum- mer):326-342. Sandell, S.H., and D. Shapiro 1978 "An Exchange: Theory of Human Capital and the Earnings of Women: A Reexamination of the Evidence." Journal of Human Resources 13, 1 (Winter):103-117. 1980 "Work Expectations, Human Capital Accu- mulation, and the Wages of Young Women." Journal of Human Resources 15, 3 (Summer):- 335-353. Stafford, F. P., and G.J. Duncan 1980 "The Use of Time and Technology by House- holds in the United States." In Research in Labor Economics, Vol. 3, edited by Ronald Ehrenberg. Greenwich, Conn.: JAI Press. Treiman, D.J., and H.I. Hartmann, eds. 1981 Women, Work, and Wages: Equal Pay forJobs of Equal Value. Washington, D.C.: National Academy Press. Weiss, Y., and R. Gronau 1981 "Expected Interruptions in Labour Force Par- ticipation and Sex-Related Differences in Earnings Growth." Review of Economic Stud- ies 58 (October):607-619. Wolfe, W., and R. Rosenfeld 1978 "Sex Structure of Jobs and Job Mobility." So- cial Forces 56:823-844. Zellner, H. 1975 "The Determinants of Occupational Segrega- tion." In Sex, Discrimination, and the Division of Labor, edited by C.B. Lloyd. New York: Columbia University Press, pp. 125-145.