Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Jobs, Job Status, and Women's Gains From Affirmative Action: implications for Comparable Worm James E. Rosenbaum Despite affirmative action and other equal employment opportunity (EEO) efforts of the 1970s, the male-female wage gap persists. The limita- tions of past efforts have made us realize that the problem may be structural in origin and may be due to sex segregation in jobs and to the way jobs are compensated. Unfortunately, most social science research on pay equity has studied individuals' earnings, which may not reveal institutional mecha- nisms that contribute to much of the pay difference among jobs in a fimn. Moreover, in a labor market in which men and women often hold entirely different jobs, the relative pay of individuals within the same jobs may be a less important source of discrimination than how different jobs are compen- sated across pay levels (Treiman and Hartmann, 19811. In particular, the issue of comparable worth poses the question of whetherjobs held predomi- nantly by women are undervalued, and thus undercompensated, relative to jobs of comparable worth that are held predominantly by men. Conse- quently, the study of pay equity must consider how the level of compensation forjobs is determined. Compensation systems in organizations often operate by a two-stage proc- ess in which the relative value of jobs is first determined and then the relative pay of individuals within jobs is determined. The latter process involves the assessment of an individual's contribution compared with the contributions of others in the same job, and this assessment usually by the supervisor of all these people- tends to operate in the way neoclassical economists cus- tomarily assume. However, the former process, determining the value of jobs, requires 116
JOBS, JOB STATUS, ED AFFI~ATWE ACTION 117 evaluating highly different tasks and classifying them into a limited number of ranked pay levels. Since these pay levels generally confer differential social status, they may be termed status categories or statuses. This process of evaluating the worth of jobs has recently become a focus of interest but has not yet received much empirical attention. The study reported in this paper provides one of the first opportunities to examine this issue empirically. The process of assigning differential status rankings to jobs initially evolved through traditional customs and beliefs. Some jobs were considered to have more status because they required more skill or were difficult (or for a diversity of other reasons), and this consensus on status was the basis for compensation differences. In recent decades these traditional status rank- ings have come under increasing criticism, and, in response, formal job evaluation systems have been devised to make the process of assigning jobs to differential hierarchical positions more objective, rational, and equitable. This change has raised the question of whether this shift from tradition to job evaluation has led to greater pay equity between jobs held predominantly by men and those held predominantly by women. This paper describes the results of a study of these issues in a large corporation. First, the study investigates how jobs are compensated, how jobs are allocated to different status categories, and the extent to which job status prevents a narrowing of the wage gap between men and women. Second, since the corporation implemented an increasingly strong aff~rma- tive action program over the period studied, the study investigates what affirmative action has accomplished in this corporation and where it falls short. Third, since job evaluation was implemented in this firm, this study provides an opportunity to observe and draw lessons concerning the use and limitations of that tool in achieving pay equity. Three findings are of particular importance. First, even a very strong aff~rrnative action program that leads to very great benefits for some women may neglect other groups of women, and, in particular, it may aid the class of women but not lead to reparation for the individuals most hurt by past discrimination. Second, affirmative action may help individuals while maintaining forms of structural discrimination against female jobs that may subsequently undermine the long-term impact of affirmative action efforts. Third, reform efforts to implement job evaluation, even in the context of a strong aff~rrnative action program, may not overcome the salary disadvan- tages of female jobs and, most perniciously, may inadvertently serve to legitimate traditional values under the mask of scientific procedure. These findings suggest that developing a comparable worth strategy may be neces- sa~y to improve women's earnings and that such a strategy, which requires changing job structures, is likely to be difficult to implement. This paper is based on the results of a larger study, which is reported in a
118 ROSENBA UM recently completed book, Career Mobility in a Corporate Hierarchy (Rosenbaum,1984~. In this brief paper statistical results are not discussed in detail. Readers wishing more detail about the analyses and theoretical con- ceptions alluded to here are referred to that book. SETTING AND DATA This paper is based on a study of employees' earnings and job attributes in a large corporation over a 13-year period from 1962 to 1975. The corpora- tion employed between 10,000 and 15,000 employees in the period studied, making it slightly larger then the median corporation in Fortune's 500 largest corporations. The corporation is a large, autonomous, investor-owned firm, having offices in many cities and towns across a large geographic region. The study had use of the computerized personnel records of the corpora- tionfortheyearsl962, 1965, 1969, 1972,andl975,permittinganalysesof changes over 3-year intervals for all but one period. The analyses reported here used two kinds of samples. One set of analyses is based on independently drawn 25 percent random samples of all individuals in the firm in four periods (the 1962 data were excluded because they lacked information on race). Regression analyses were used to explain individuals' earnings (in logs, adjusted to 1967 dol- lars), in terms of individual attributes (experience, education, entry age, sex, race, and the status ranking of colleges attended [using the Astin 1965 scaler) and comparing their relative influences within and across job status levels. The other set of analyses examined the determinants of the average pay that jobs offer. These analyses are based on a data set that aggregates infor- mation about all individuals in each job (i.e., a 100 percent sample) to describe jobs in terms of averages: average tenure, average earnings, per- centage of college graduates, percentage female, and so forth. These analy- ses took jobs as the units of analysis in order to discover relationships among the attributes of jobs. The same four time periods are analyzed. The data for these analyses were taken from the computerized file of the complete personnel records of the corporation. The availability of actual personnel records makes this research distinct from most economic and sociological research in this area and offers two important advantages. First, these personnel records offer valid and precise information on employees' salaries, levels of attainment in the organizational hierarchy, education, college attended, age, and years of experience in the corporation (tenure). Most research in this area has been based on employees' self-reports, and the possibilities of distortions or poor recollection are generally a cause of concern. In contrast, these personnel records are systematically kept by the
JOBS, JOB STATUS, AND AFFIRMATIVE ACTION ~ ~ t ~ 119 co~pora~on-s personnel Department tor me organization's own use; conse- quently, extensive data-checking procedures have been implemented to ensure the accuracy of the information, since mistakes would have impor- tant implications for company personnel practices. Second, these personnel records represent the complete universe of employees in the corporation. Most studies of organizations have had to rely on employees' cooperation, and less-than-perfect cooperation leads to incomplete information or possibly biased samples. For instance, Grusky's (1966) 75 percent response rate (in the data used in Halaby, 1979) is rather good as questionnaire response rates go; however, it does raise questions about the representativeness of the sample that cannot be completely settled. The data in Rosenbaum (1984) were drawn from the universe of all employ- ees in the cohorts studied, and no information is missing (except the above- noted race information in 19624. Of course, the use of personnel records does have its own limitations, primarily in that they do not include all the information that a researcher might like to consider. In particular, the records do not include information on marital status or actual years of education, which previous research has suggested to be worthy of consideration. EllECTS OF MATH ACTION ON CHUMS This organization implemented a serious affirmative action program in 1970 and an even stronger one in 1973, and these two programs had an impressive beneficial impact on women's earnings. In 1969, before these programs began, women received about 46 percent lower earnings than men with similar characteristics at entry, and their disadvantage remained at about the same magnitude for those with greater seniority. The first and second affirmative action programs increasingly reduced the female disad- vantage: from 46 percent in 1969 to 39 percent in 1972 and 23 percent in 1975. Female college graduates gained even more, going from a 36 percent disadvantage for new entrants before affirmative action to only a 3 percent disadvantage by 1975. However, more-senior women did not fully share in these gains. While seniority benefited women et the same rate as men in 1969, by 1972 their rate of earnings increase with seniority was 21 percent lower than that of men, and by 1975 it was 29 percent less. For example, in terms of the magnitude of their earnings, entering women in 1975 had a 23 percent earnings disad- vantage relative to male peers, but women with just 10 years of seniority had a 31 percent disadvantage relative to male peers. At 20 years' tenure the disadvantage was even greater. Obviously the great gains for newly entering women were not being fully shared with their more senior peers.
120 ROSENBAUM This outcome is particularly poignant, because it was the more-senior women who had been most hurt by the previous decades of discrimination and whose experience had promoted the affirmative action program. None- theless, the affirmative action program was not providing reparation for the deprivation these women had experienced. The specific individuals who had been most hurt by previous discrimination were left with earnings similar to what they had always been. As observers of the civil rights movement have noted, reforms have differential impact on various members of a deprived group. What is the reason for this discrepancy? Although individual attributes may be involved, analyses also indicate a social structural mechanism that mediated most of this discrepancy. Most of the increasing discrepancy for more-senior females was mediated by job status in the organizational hierarchy. Specifi- cally, when job status was introduced into the earnings regression, it accounted for most of the negative effect of tenure for women (see Rosen- baum, 1984:Ch.71. Further analyses indicate that newly entering women received great increases in job status from the affirmative action program, but more-senior women continued in the same jobs or in jobs of similar low status. The job status system created an important obstacle to earnings gains ~ . for senior women. Having found that job status played a role in preventing senior women from sharing in the benefits of affirmative action, the following analyses seek to study how jobs and the composition of jobs may contribute to this process. THE CREATION AND DISSOLUTION OF JOBS For the purposes ofthese analyses, we must define the sample of jobs to be analyzed and consider to what extent jobs are stable entities in this corpora- tion over time. Of course, this issue has very practical implications. If the same set of jobs stably exists over time, then a one-time comparable worm reassessment will have enduring effects, while if jobs are continually being created and disbanded, then comparable worth efforts must be made on a continuing basis. Large organizations are usually conceptualized as unchanging, almost physical structures much like the concrete and steel structures that house them. Organizational charts of jobs in an organization are portrayed as if these diagrams represented something enduring about the organization. Yet organizations often change their organizational structures; jobs are created and destroyed. Technological and economic changes occur, requir- ing new kinds of jobs and making others obsolete. Political battles over authority and task responsibilities change jobs into entirely different jobs.
JOBS, JOB STATUS, AND AFFIRMATIVE ACTION 121 Individual job occupants become proficient and take on so many additional tasks that they change the nature of a particular job. These dynamic pro- cesses seem entirely plausible, but we have no idea how often they occur. To the extent that they occur, they pose a serious challenge to our notion of a fixed organizational structure, for they suggest that old job duties or old ways of organizing duties into jobs disappear and new duties or new ways of organizing duties appear in the form of new jobs. If jobs and job statuses are as important as they appear to be in affecting earnings, this degree of change in the job structure will have important implications for any program designed to affect earnings and employment outcomes. The study analyzed the set of job titles listed in the personnel records for the years 1962, 1965, 1969, 1972, and 1975. Titles for the three lowest levels in the organization nonmanagement, foreman, and lower manage- ment are considered in this paper. The job title identifications for higher levels were removed from the data to protect the anonymity of employees, since these jobs tended to have few occupants. Their small size also made them unsuitable for analysis. The main concern of the first analysis was the number of jobs that appeared or disappeared, i.e., that went from having no occupants to having some occupants, and vice versa. However, since jobs with few occupants could mistakenly appear to have been eliminated when they temporarily have several vacancies, inferences are more ambiguous with these jobs. Consequently, the analysis distinguished between small jobs (having one to four occupants) and large jobs (having five or more). We can be more confident in inferring that jobs have been eliminated when jobs with five or more occupants subsequently have no occupants. The pattern of changes in job occupancy is one of high levels of job disappearance and creation. For example, in nonmanagement, 26 of the 95 small jobs (with one to four occupants3 in 1962 had become empty 7 years later, and 4 of 76 large jobs (with five or more occupants) in 1962 had become empty 7 years later. Even more new jobs were created over these time intervals. Between 1962 and 1969, 26 small and 14 large nonmanage- ment jobs were created. The patterns for other jobs show even greater amounts of job disappear- anceandjob creation. While 67 percept ofthe nonmanagementjobs existing in either 1962 or 1969 were existing in both years, only 54 percent of the foreman jobs and 41 percent of the lower-management jobs were similarly enduring. For the 13-year interval between 1962 and 1975, the comparable figures were much lower for all jobs and the difference among jobs was even greater: 53 percent for nonmanagement, 46 percent for foreman, and 22 percent for lower-management jobs. For all three kinds of jobs, most jobs existing in these years were abolished or newly created over the period; persistence is more the exception than the rule.
122 ROSENBA UM In considering the possible generalizability of these findings, what is most striking is that this company has not undergone enormous outwardly visible changes. It continued to conduct the same activities over the entire period. Nor have there been any mergers with other companies nor divestitures of subsidiaries. To all outward appearances, it is the same organization housed in most of the same buildings doing largely the same activities over the entire decade. Nor has it had any new leadership or organizational shake-ups that would explain these changes. As large corporations go, it is probably more stable and less changed than most over the period considered. Of course, affirmative action, job evaluation, organizational contraction, and organizational growth were important changes over these periods, but they are not sufficient to explain the changes observed. The 1962-1969 period in particular is largely unexplained by these changes, since most of them occurred after this period. Organizational growth is the main feature of the period, and, although growth may explain job creation, it does not explain why more jobs were disbanded than were created over the period. The disappearance and creation of jobs found here are indicative of small changes that gradually evolved from a multitude of particular decisions in various parts of the organization. Although no other evidence is available on this specific point, it is reminiscent of Granovetter's finding that 35 percent of the job placements in his study were jobs that had not existed previously (Granovetter, 1974:141. These results have important implications for job evaluation programs. Although a jobs-based policy may be necessary for realizing goals for afh~r- mative action and is essential in developing a comparable worth strategy, a lasting solution will not result from a one-time effort. With jobs continually being disbanded and created, any reform effort requires the formulation of an entirely new system and a continuing commitment to its implementation. This analysis also clearly indicates that our customary image of a fixed organizational hierarchy is seriously misleading. Jobs appear and disappear at a surprisingly high rate. White's (1970) procedure for analyzing mobility in terms of vacancy chains would clearly not apply here, even over the short period 1962 to 1969. Although the assumptions of vacancy analysis seemed to apply rather well to the hierarchy of the Episcopal church that White studied, in the corporation studied here any vacancy that occurred could go unfilled forever, while new jobs were being created and filled that had never before been occupied. DO JOBS OllER ST - LE CO~ENSAMON? To what extent do jobs change in the relative compensation they offer over periods of changing economic circumstances? The importance of jobs and
JOBS, JOB STATUS, AND AFFIRMATIVE ACTION 123 job statuses to earnings will be especially great if it is found that jobs provide stable relative earnings over time. We commonly assume that jobs retain the same value over time, so that compensation would not be very responsive to changes in economic circum- stances. However, these hypotheses are put to a severe test in studying the periods included in these data, for they were times of economic and social change. The successive periods in these data were times of modest growth (1962 to 1965) followed by booming growth (1965 to 19691. We would expect such changes to be accompanied by changes in priorities among types of jobs. Furthermore, the organization shifted from a traditional job status system to a job evaluation system during this period (in 19721. In addition, an affirmative action program begun in 1970 was made even stronger in 1973 developments that we would expect to change the relative standing of jobs filled with large proportions of women and minorities. In the face of these great changes, to what extent did jobs change in the compensation they offered? The job titles data file was created from the complete personnel records on all employees in the firm. These data on individuals were aggregated by job titles, so that jobs became the units of analysis, and the variables for these units were the average salary of all occupants in each job, the average tenure of job occupants, the percentage of job occupants with a college degree, the percentage of job occupants who were female, and so forth. This section reports the analysis of the compensation offered by a job. This is indicated by the average earnings of all individuals who occupy the job at a particular time and is called job salary or simply salad. There is a large range of job salaries in this organization, even within a single level of the organizational hierarchy. ' To what extent are these job salaries stable properties of jobs over these periods of great economic changes? The analysis finds only minimal change in salaries. Job salaries in non- management levels remained extremely stable between 1962 and 1975 (Table 1~. All correlations are large: the correlations of 1962 job salaries with those in later years show negligible decline (.98, .97, .96, .961. The comparable correlations at the foreman level start at the same magnitude, but the correlations of 1962 job salaries with those in later years show some decline, particularly between 1969 and 1972 (.96, .94, .88, .841. (In this and l In the year 19629 the job salaries of nonmanagement jobs ranged from $3,354 to $8,684, with a mean of $5,462 and a standard deviation of $1,336 for the 171 jobs on that level. Foreman jobs offered job salaries in the range of $4,810 to $10,061, with a mean of $8,031 and a standard deviation of $1,511 for the 155 jobs. Lower-management job salaries were between $9,222 and $12,203 with a mean of $1 1 ,422 and a standard deviation of $729 for the 86 jobs.
124 TABLE 1 Stability of Job Salaries by Level (Pearson correlations) ROSENBA UM Nonmanagement Level 962 1965 1969 1972 965 .977 1969 .968 .980 1972 .958 .966 .968 1975 .958 .954 .948 .954 Foreman Level 1965 1969 1972 1975 1962 .965 .944 .876 .842 1965 .952 .881 .868 1969 .920 .869 1972 .898 subsequent analyses, the lower-management level was not analyzed, because too few jobs persist over these periods.) Despite considerable changes in economic circumstances and organiza- tional practices over these periods, job salaries remained extremely stable for nonmanagement and foreman levels. This stability is particularly strik- ing, for it occurred despite an explicit effort to reform the traditional job status system by implementing job evaluation at foreman levels and above. This effort was implemented in 1972 and was probably responsible for the declining earnings correlation between 1969 and 1972 at the foreman level. However, the decline in the earnings correlation is surprisingly minor, given the extensiveness of the reform. This stability suggests that the job evalua- tion effort had little effect on the average earnings in jobs. THE SEX COMPOSITION OF JOBS ANI) WOMEN'S EARNINGS Having seen that jobs offered highly stable salaries over this period, our analysis of the effects on women's earnings must turn to an analysis of the composition of jobs to understand how women are allocated among jobs and how this affects their earnings. We begin by describing the sex composition of jobs and how it was affected by affirmative action. The affirmative action program had an increasingly strong effect in inte- grating jobs. Of the 100 nonmanagement jobs with five or more individuals in 1965, only 6 contained both men and women (Table 21. Almost 40 were all-female jobs and 55 were all-male jobs. The numbers remained similar in
JOBS, JOB STATUS, AND AFFIRMATIVE ACTION TABLE 2 Jobs at Each Level by Percentage Female, 1962-1975 (forjobs with five or more employees) 125 Level 1962 1965 1969 1972 1975 Nonmanagement O percent 47 55 42 34 19 1-99 percent 3 6 7 13 25 100 percent 2 37 34 31 31 Total 73 98 83 78 75 Foreman O percent 18 25 20 17 17 1-99 percent 2 7 7 13 14 lOO percept 24 27 25 18 20 Total 44 59 52 48 51 Lower-management O percent 13 21 18 9 8 1-99 percent 0 0 5 4 8 100 percent _ O 0 1 1 Total 13 21 23 14 17 the following period, but by 1972 the modest affirmative action program had doubled the number of sex-integrated jobs (to 13), and the number doubled again by 1975 (to 251. Similar changes occurred in the foreman and lower- management levels. The affirmative action program seems to have had great success in increasing the number of integrated jobs . However, despite these gains, most jobs were unaffected. By 1975, even with a generous definition of integration as the broadest possible range (1 to 99 percent women), only one-third of nonmanagement jobs fit into this range, and only 27 percent of foreman jobs did so. Clearly, the vast majority of jobs were still segregated by sex. Moreover, the sex-segregated jobs offered considerably different salaries. All-female jobs paid an average salary that was about three-quarters of the salary paid by all-male jobs, and this ratio applies both to nonmanagement and foreman levels in most years (Table 31. There is a general trend toward slightly greater equality over these years, but the trend is modest (and it shows an unexplainable reversal at the nonmanagement level in 19751. Female jobs clearly showed much less gain than individual women experi- enced over this period. Of course, the simple association between female composition and aver- age salary may be highly misleading, since jobs may differ in other respects
126 ROSENBA UM TABLE 3 Average Salary forJobs by Percentage Female in Jobs, 1962-1975 (forjobs with five or more employees) Level 1962 1965 1969 1972 1975 No n man ageme n t O percent $ 6,268 $ 6,988 $ 7,894 $10,009 $14,679 100 percent 4,530 4,854 5,850 7,590 10,632 Foreman O percent 8,863 10,050 12,345 15,297 19,713 100 percent 6,849 7,440 9,104 11,569 15,722 Lower-management O percent 11,160 12,440 14,617 19,368 22,600 100 percent 18,333 21,985 and the apparent effects of female composition may actually be due to other factors. For instance, jobs may differ in terms of the education and training they require, and economists might suggest that the human capital composi- tion of jobs may account for the apparent effects of female composition. Regressions were nan to study the effects of female and minority composi- tion on job salaries, controlling for the percentage of college graduates in the job and the average tenure in the job (Table 4) . As economists would predict, education and training affected job salaries, but percentage minority and percentage female also had significant independent effects. Percentage female had a strong significant negative relationship with job salaries in all years, even after controlling for human capital composition. At the nonmanagement level in 1965, jobs were paid $21 less for each addi- tional female percentage unit, so that the regression estimates that jobs with 100 percent females were paid $2,100 less than all-male jobs, a figure very close to the real difference between the job salaries of all-male and all- female jobs. Although this difference declined over the periods of increas- ingly strong affirmative action, even in 1975 it retained a strong significant negative effect. At the foreman level the decline was less, and it was more erratic (perhaps because of the large influx of women promoted into fore- man-level female jobs in 19721. But even in 1975 female composition still had a strong negative association with job salaries. Although human capital composition accounts for some of the earnings differences among jobs (particularly at the foreman level), even after controlling for this, jobs with a higher percentage female offered lower salaries.2 2 The large unexplained changes in the influence of minority percentage may be due to the small numbers of minorities in most jobs.
JOBS, JOB STATUS, AND AFFIRMATIVE ACTION TABLE 4 Human Capital, Sex, and Race Effects on Job Salaries Over Four Periods 127 Level 1965 1969 1972 1975 Nonmanagement Percentage college 27.814* 22.313* 25.821* 26.522* graduates (5.175) (5.173) (6.012) (6.842) Tenure .319* - .387* .898* .586* (.099) (.097) (.227) (.259) Percentage minority 38.045* -22.667* - 14.207 - 24.159* (14.112) (7.453) (7.898) (8.749) Percentage female 20.759* - 19.868* - 14.946* - 12.310* (4.383) (5.192) (5.170) (5.031) Constant 75.423* 81.330* 66.212* 75.710* (2.874) (3.290) (3.640) (4.473) Variance explained (R2) 43.1 % 39.3 % 40.3 % 37.3 % Standardized coefficients Percentage college graduates .416 .401 .390 .370 Tenure .248 - .376 .382 .231 Percentage minority .209 - .276 - .172 - .278 Percentage female - .369 - .339 - .258 - .231 Foreman Percentage college 7.116* 10.718* 12.783* 11.022* graduates (1.906) (3.550) (3.907) (3.495) Tenure .256* .046 .512* .460* (.103) (.120) (.192) (.195) Percentage minority -2.623 -84.099* -9.941 - 1.147 (13.479) (22.241) (6.320) (5.737) Percentage female - 10.259* - 8.108* - 10.200* - 8.192* (1.122) (1.011) (1.417) (1.460) Constant 95.278* 103.558* 103.885* 104.559* (2.519) (2.966) (4.538) (4.841) Vanance explained (R2) 40.0% 46.7% 41.5% 32.4% Standardized coefficients Percentage college graduates .224 .203 .262 .283 Tenure .150 .025 .219 .212 Percentage minority - .011 - .246 - .121 - .017 Percentage female .560 - .534 - .541 - .472 * Coefficient significant at .05 level.
128 ROSENBAUM TABLE 5 Human Capital, Sex, Race, and Status Effects on Job Salaries Over Four Periods Foreman Level 1965 1969 1972 1975 Percentage college -2.559* -2.781 .509 - 1.689 graduates (.571) (1.529) (2.236) (1.381) Tenure .111* - .079 .437* .175* (~029) (.049) (.103) (.073) Percentage minority 7.646* - 26.520* - 2.172 .053 (3.749) (9~269) (3.435) (2.115) Percentage female - .778* - 1.681 -2.870* - .250 (.376) (.480) (.887) (.621) Job status 3.169* 2.989* 9.164* 9.502* (.070) (.117) (.567) (.373) Constant 61.949* 72.561* 69.731* 73.949* (1.017) (1.704) (3.230) (2.151) Variance explained (R2) 95.4% 91.4% 83.2% 90.9% Standardized coefficients Percentage college graduates - .081 - .053 .010 - .043 Tenure .065 .043 .186 .081 Percentage minority .035 - .078 - .026 .001 Percentage female - .042 .111 - .152 - .014 Job status .973 .894 .812 .949 n= Partitioning of variance from Table 5 Variance (%) 44 59 53 52 Unique Total Unique Total Unique Total Unique Total Demographic variables 1.5 40.0 2.1 46.7 5.2 41.5 1.2 32.4 Status 55.4 93.9 44.7 89.3 41.7 78.0 58.5 89.7 Shared 38.5 44.6 36.3 31.2 Total 95.4 91.4 83.2 90.9 * Coefficient significant at .05 level.
JOBS, JOB STATUS, AND AFFIRMATIVE ACTION 129 Moreover, further analysis indicates that the job status system was involved in mediating much of the effect of female percentage. Adding job status to this regression reduces the net effect of percentage female a great deal (Table 51. Forinstance, while each additional unit of female percentage reduces job salaries at the nonmanagement level by $19.87 in 1969, each female percentage unit subtracts only $1.68 after controlling for job status. This indicates that nearly all of the effect of female percentage on job salaries is due to the lower statuses of these jobs. What is most noteworthy is that this remains true even after the organization shifted from a status system based on tradition to a system based onjob evaluation. In 1972, afterjob evaluation was implemented, each female percentage unit subtracts almost $15, but after controlling for job status, each female percentage unit subtracts less than $3. Job evaluation created a status hierarchy that continues to mediate much of the effect of female composition on earnings in the same way that the traditional status system did. Job evaluation, even in the context of a strong affirmative action program, maintains and possibly reinforces much of the earnings differences among male and female jobs. PRO M OTION CHALICES ASSOCIATED VVITH JOBS This analysis turns to another issue that is sometimes ignored in the discussion of job rewards. Jobs differ in the promotion opportunities they offer. Human capital economists might attribute this to the differential train- ing offered by jobs, while structural sociologists would explain it in terms of opportunity structures. Regardless of cause, however, these data show that the percentage of individuals who are promoted from a given job tends to be highly stable. The percentage of employees promoted from 1962 jobs in the next 3 years is highly correlated with the percentage promoted from the same jobs a decade later (r = .807 at the foreman level. .735 at the nonmana~e- ~ _ _ _ _ _ ~ ment level). ., _ , ., __ ~ - .,_ ^,_,,~,,~,,~e,_ Moreover, jobs have lasting effects on individuals' careers. The 1975 attainments of the cohort of employees who entered this organization between 1960 and 1962 were analyzed (Rosenbaum, 1984:Ch. 6~. Using separate (dummy) variables for each job, the longitudinal analysis indicates that individuals' 1962 jobs had a veer large effect on their attainments 13 years later, even after controlling for sex, race, age, education, and a rough proxy measure of ability (the Astin 1965 scale of college quality). As an indication of unmeasured components of human capital, 1962 earnings were also introduced into the regression. In all analyses, early jobs continued to have a strong and significant effect on attainment 13 years later (see Rosen- baum, 1984:Ch. 6~. Moreover, this effect does more than simply advance an individual to a
130 ROSENBAUM higher job in the following period. Even controlling for intervening attain- ments (in 1965), early jobs had a significant enduring effect on individuals' subsequent careers. Even employees who did not advance in their first few years still benefited from early placements in higherjobs. According to the Horatio Alger stories that are often cited in this organization, individuals who overcome initial low origins are offered advancements to the highest echelons. Analyses found that some of these individuals (who entered the firm in the years 1960 to 1962) did indeed overcome low origins by 1965 to -I attain the same jobs as "elites" who had been in these higher-status jobs from the outset. When their attainments 10 years later (1975) were com- pared, however, the Horatio Alger types ended up much lower then the elites who had been initially favored with betterjobs. Early jobs have an enduring effect that even subsequent attainments do not eradicate. Moreover, most of the effect of individual jobs was due to the way jobs are ranked in the job status system. When these analyses were repeated replac- ing the job dummy variables by the single job status variable, the results remained quite similar. Early job status explained nearly as much variation in later job status (or later earnings) as did dummy variables for all the individual jobs. Clearly, the job status system in this company has an impor- tant impact on employees' careers. Given the great importance of jobs and job statuses in determining employees' careers, it is noteworthy that the promotions associated with jobs are often ignored as job effects. The reason is clear; it is not easy to determine the effects of jobs on future promotions in short-run analyses. It is clear, however, that the ultimate impact of jobs will be related to these kinds of outcomes. The data on this firm permit an analysis of these issues. Taking as the dependent variable the percentage of employees promoted from a job in the next 3 (or 4) years, regression analyses were run using the same model as was used to explainjob salaries. Controlling for percentage of college gradu- ates and average tenure, the regressions found that the female concentration of a job was negatively related to the promotion chances of a job during the 1960s (Table 61. However, the affirmative action program had a strong effect on this result. Over the two periods in the 1970s, the negative influ- ence of percentage female continually decreased at the nonmanagement level, and at the foreman level its influence vanished in the 1969-1972 period and actually became significantly positive in the final period. More- over, when job status was added to this regression, the effect of percentage female on 1969-1972 promotions declined slightly and had no effect at all on 1972-1975 promotions (Table 7~. In spite of the high stability of promotion rates offered by jobs, the stability is not total, and jobs with high proportions of women experienced a large reduction in their disadvantage. These results may be interpreted as reassuring in some respects. While
JOBS,JOBSTATUS,ANDAFFIRMATlVEACTlON 131 TABLE 6 Human Capital, Sex, and Race Effects on the Promotion Chances of Jobs Over Four Periods Level 1962-1965 1965-1969 1969-1972 1972-1975 No n manage me nt Percentage college .688* 1.167* .584* .697* graduates (.095) (.223) (.171) (.091) Tenure .005 - .002 - .004 - .001 (.005) (.004) (.003) (.003) Percentage minority .823 .172 - .096 (.606) (.247) (.122) Percentage female - .108 - .321 - .140 - .114 (.169) (.218) (.184) (.084) Constant .141 .380* .351 * .114* (.078) (.123) (.109) (.056) Variance explained (R2) 42.7% 24.7% 12.9% 45.3% Standardized coefficients Percentage college graduates .655 .466 .382 .665 Tenure .083 - .050 - .132 - .036 Percentage minority .121 .075 - .073 Percentage female - .057 - .131 - .081 - .117 n= 76 101 85 81 Foreman Percentage college graduates Tenure Percentage minority Percentage female Constant Variance explained (R2) Standardized coefficients Percentage college graduates Tenure Percentage minority Percentage female .439* .491* ( 105) (.079) .001 .001 (.003) -.085 (.071) .029 (.086) 39.0% .591 .023 .154 (.004) - 1.368 (1.401) -.136 (.082) .129 (.096) 50.6% .650 .015 -.095 .369* .104) .000 (.003) .419 (.314) .010 (.065) .065 (.067) 35.6% .473* (.034) -.004 (.003) -.432* (.185) .069* (.032) .084 (.056) 86.1% .888 .010 -.095 -.143 - .165 .019 .120 n= 43 59 53 50 * Coefficient significant at .05 level.
132 ROSENBAUM percentage female and job status remained serious obstacles to earnings parity, they did not limit women's promotions after the affirmative action program began. Jobs with high proportions of women had as high or higher promotion rates as jobs with high proportions of men. However, although these specific results are gratifying, some caution is required. The high rates of female promotions that were demanded by the organization's affirmative action program are unlikely to be maintained after the organization reaches its initial targets and the impetus for the affirmative action program subsides. At that time it may be expected that the organiza- tion will go back to more selectivity in its promotions of women, and then the customary reliance on job status for promotions will once again hurt individ- uals in the low-ranked, predominantly female jobs. Although the promotion findings are reassuring in the short run, even these findings call for some concern about how the job evaluation system is ranking female jobs. Our awareness of the traditional historical relationship between job status and promotions suggests that one must be wary about whether the new pattern will endure. CONCLUSIONS This study sought to discover how jobs and job statuses affect women's earnings and how they affect women's gains from an affirmative action program. At the individual level the analyses indicate that job status is an important determinant of women's earnings. The job status system deter- mined which women benefit and which do not benefit from the strong affirmative action program. In particular, the job status system prevented more-senior women from gaining as much from the affirmative action pro- gram as their less-senior peers. In order to get another perspective on job effects, the properties of jobs were investigated. The continual appearance of new jobs clearly suggests that lasting reforms will not result from one-time efforts. Reforms must be embodied in coherent systems, like job evaluation systems, that can con- tinue to handle the many new jobs that are created. The findings also indicate the need for revaluing the worth of jobs. In the 1960s the female composition of jobs had a strong negative influence on job salaries. While previous research has shown such a relationship for occupa- tional groups (Sanborn, 1964; Fuchs, 1971; Oaxaca, 1973; Sommers, 1974; Treiman and Terrell, 1975; Featherman and Hauser, 1976; Blau, 1977; Treiman et al., 1984; Roos, 1981), it has not been possible to study it for jobs within firms. The finding of female job composition effects on job salaries is particularly noteworthy because these analyses control not only for human capital composition but also for level in the authority hierarchy.
JOBS,JOBSTATUS,AND AFFIRMATIVE ACTION TABLE 7 Human Capital, Sex, Race, and Status Effects on the Promotion Chances of Jobs Over Four Periods 133 Foreman Level 1962- 1965 1965- 1969 1969- 1972 1972- 1975 Percentage college .332* .347* .280* .456* graduates (.109) (.081) (.105) (.039) Tenure .002 - .001 .000 - .004 ( 003) (.004) ( 003) (.003) Percentage minority - .894 .049 - .362 (1.271) (.352) (.199) Percentage female - .005 - .027 .092 .077* (.074) (.080) (.070) (.033) Job status .014* .021 * .014* .008 (.006) (.006) (.006) (.009) Constant .053 - .039 - .095 .054 (.088) (.100) (.090) (.064) Variance explained (R') 47.1 ~ 60.8 ~ 43.2 5'0 86.4 Standardized coefficients Percentage college .444 .460 .422 .857 graduates Tenure -.077 -.032 -.009 -.100 Percentage minority - .063 .019 - .120 Percentage female - .009 .032 .166 .136 Job status n= .344 .401 .391 .064 43 59 53 50 * Coefficient significant at .05 level. Moreover, the job status hierarchy mediates most of the effect of female composition on job salaries. Clearly, some revision of the job status system is necessary if women's jobs are to be better paid. However, job evaluation does not necessarily contribute to gains for women. The findings indicate that job evaluation, even in the context of a strong affirmative action program, does not diminish the relationship between female composition and job salary veer much. Indeed, the new job
134 ROSENBAUM status system based on job evaluationcontinued to mediate this relation- ship. Two kinds of cautions need to be borne in mind in evaluating these findings. First, the evidence is not sufficient to identify unambiguously whether the effect is indeed discrimination. That requires far more detailed data than this research, like most research, encompasses. However, the discrepancy between the gains of some individual women and the lack of gains for other women and for female jobs in general suggests that structural barriers may be operating, although other interpretations are possible. Second, since this is a case study of only a single corporation, it is difficult to assess the generalizability of most of these findings. The initial finding of this study that job status mediates most of the female earnings disadvan- tage has been shown in studies of other organizations (Malkiel and Malkiel, 1973; Halaby, 1979), but its effect in preventing senior women from gaining from affirmative action and in mediating the influence of female composition requires further replication. Similarly, the effect of female composition on salaries has previously been shown for occupations but not forjobs, and it has not been possible before to relate it so clearly to job evaluation status categories. However, regardless of generalizability, the primary value of these find- ings is in suggesting some issues to consider in assessing the effectiveness of affirmative action programs and in developing comparable worth strategies. Three central conclusions are stressed. First, even a very strong affirmative action program that leads to very great benefits for some women may neglect other groups of women. In particular, aiding women as a class may not lead to reparation for those most hurt by past discrimination. Second, to the extent that these programs are based on helping individual women, as most such programs are, they are likely to have the problem observed, namely, helping some individual women while maintaining struc- tural discrimination against female jobs. Not only does this prevent the individuals in these jobs from receiving the full benefit of the affirmative action program, but also, more seriously, it remains a structural component of the organization so that when the impetus of the aff~',~ative action pro- gram ends, the structure may again tend to recreate old patterns. Third, reform efforts to implement job evaluation, even in the context of a strong affirmative action program, may not overcome the salary disadvan- tages of female jobs. The conclusion to the interim report of the National Research Council's Committee on Occupational Classification and Analysis warned that job evaluation systems use several procedures that raise ques- tions about whether they can be effective in fairly assessing sex-segregated jobs (Treiman, 1979:481. In particular, their reliance on market wage rates and subjectivejudgments raises serious risks that job evaluation will recreate
JOBS, JOB STATUS, AND AFFIRMATI VE A CTION 135 the same biases as traditional status systems. Unfortunately, the present findings provide empirical support for this warning, and they suggest that job evaluations may be especially pernicious in providing legitimation for traditional values under the mask of scientific procedure. Obviously, the ultimate answer to this problem is to integrate all sex- segregated jobs. This is a noble goal, but it is one that seems to proceed slowly. Short of that ideal, the solutions are more difficult. Presumably, subjective job assessments and market-based factor weight- ings had important influences on these findings, and these may be the best targets for policy change. We currently lack sufficient evidence on how these procedures operate. However, the present findings suggest that the mechanisms underlying job evaluation need to be scrutinized to discover whether the inequalities are inequitable, and, if so, how job evaluations could be done in other ways to lead to more equitable results. In the absence of data for analyzing the job evaluation process in detail, this study offers empirical support to the warning of the National Research Council report: jobs and job evaluation programs, even in the context of very serious and effective affirmative action programs, may partially undermine the goals of affirmative action and legitimize these inequalities. Given the difficulty of criticizing job evaluation programs, this may be a serious obsta- cle to preserving panty after the impetus for aff~rrnative action programs is reduced. It also suggests caution in relying on "unreformed" job evaluation plans in comparable worth strategies. Those seeking a realignment of the wage rates of women's jobs should attempt to ensure that the job evaluation instrument used is not unfairly biased toward maintaining the status quo. ACKNOWLEDGMENT The author is indebted to the conference participants particularly Bill Bielby, Heidi Hartmann, and Don Treiman - for their helpful suggestions. Preparation of this paper was supported by the Center for Urban Affairs and Policy Research at Northwestern University and by the Russell Sage Foun- dation. The views presented here are, of course, those of the author. REFERENCES Blau, Francine D. 1977 Equal Pay in the Ounce Lexington, Mass.: Lexington Books. Featherman, David L., and Robert M. Hauser 1976 Sexual inequalities and socioeconomic achievement in the U.S., 1962-1973. American Sociological Review 41 (June): 462-483. Fuchs, Victor 1971 Differences in hourly earning between men and women. Monthly Labor Review 94(May):9- 15.
136 ROSENBA UM Granovetter, Mark 1974 Getting a Job: A Study of Contacts and Careers. Cambridge, Mass.: Harvard University Press. Musky, Oscar 1966 Career mobility and organizational commitment. Administrative Science Quarterly 10:489-502. Halaby, Charles N. 1979 Sexual inequality in the workplace: An employer-specific analysis of pay differences. Social Science Research 8:79-104. Malkiel, Burton G., and Judith A. Malkiel 1973 Male-female differentials in professional employment. American Economic Review 63:693-70S. Oaxaca, Ronald 1973 Sex discrimination in wages. Pp. 124- 151 in Orley Ashenfelter and Albert Rees' eds., Discrimination in LaborMarkets. Princeton, N.J.: Princeton University Press. Roos, Patricia A. 1981 Sex stratification in the workplace: Male-female differences in economic returns to occupation. Social Science Research 10:(3). Rosenbaum, James E. 1984 Career Mobility in a Corporate Hierarchy. New York: Academic Press. Sanborn, H. 1964 Pay differences between men and women. Industrial and Labor Relations Review 17(July):534-550. Sommers, Dixie 1974 Occupational rankings for men and women by earnings. Monthly Labor Review 97 (August): 34-51. Treiman, Donald J. 1979 Job Evaluation: An Analytic Review. Committee on Occupational Classification and Analysis. Washington, D.C.: National Academy of Sciences. Treiman, Donald J., and Heidi I. Hartmann, eds. 1981 Women, Work, and Wages: Equal Payfor Jobs of Equal Value. Committee on Occupa- tional Classification and Analysis. Washington, D.C.: National Academy Press. Treiman, Donald J., and Kell.~it Terrell 1975 Women, work, and wagesTrends in the female occupational structure since 1940. Pp. 157-200 in Kenneth C. Land and Seymour Spilerman, eds., Social indicator Models. New York: Russell Sage Foundation. Treiman, Donald J., Heidi I. Hartmann, and Patricia A. Roos 1984 Assessing pay discrimination using national data. In Helen Remick, ea., Comparable Worth and Wage Discrimination: Technical Possibilities and Political Realities. Phila- delphia: Temple University Press. White, Harrison C. 1970 Chains of Opportunity. Cambridge, Mass.: Harvard University Press.