This file was prepared for electronic distribution by the inforM staff. Questions or comments should be directed to inform-editor@umail.umd.edu. The Economic Status of Black Women: An Exploratory Investigation Staff Report United States Commission on Civil Rights October 1990 Part II The Black-White Female Wage Differential: Reasons for the Black-White Wage Gap and Sources of the Increase in Black Women's Relative Wages Chapter 3 Determinants of Black-White Female Wage Differentials: A Conceptual Framework Black women have historically earned lower wages than white women. This chapter lays out various possible explanations for why black women and white women might receive different wages. The chapter provides a conceptual framework for the study of black-white female wage differences carried out in the rest of part II. A gap between the wages of black and white women can arise from two distinct, but not mutually exclusive, sources: racial differences in productivity (or skills) and racial differences in the degree of discrimination suffered in the marketplace. Each of these factors can independently cause black women to earn less than white women. Interactions between these two factors could further lower the relative wages of black women. Human Capital Theory: Productivity and Wages According to a well-established economic doctrine called "human capital theory," workers' wages are related to their investments in "human capital," or skills that raise their labor productivity. Workers who make larger human capital investments receive higher pay than other workers. An important example of a productivity-increasing human capital investment is a worker's investment in education. Typically, the more education the worker has, the more productive the worker is, and the higher the wage received. Another common form of human capital investment is on-the-job training, or time spent learning skills on the job. Because many skills are acquired on the job, workers with more work experience generally receive higher wages. Even if there were no employment or wage discrimination against black women, black women could receive lower pay than white women if they made smaller human capital investments. What, then, could lead black women to make smaller human capital investments than white women? One reason black women might acquire less education and other forms of human capital than white women is that they typically come from poorer families. Because human capital investments are costly, low-income people generally acquire less human capital than high-income people.(1) Thus, black women might be expected to invest less in human capital on average than white women because of family income differences. On the other hand, there are strong reasons to believe that black women might choose greater human capital investments than white women. The higher labor force participation rates of black women mean that the typical black female worker probably spends a larger portion of her life working than the typical white female worker and thus has a longer period for her human capital investments to pay off. As for any other form of investment, the longer the payoff period, the greater the amount invested. To the extent that black women anticipate spending more years working than white women, black women would be expected to make larger human capital investments, both in education and in on-the-job training. Becker's Theory of Discrimination A well-known economic theory developed by Gary Becker suggests that labor market discrimination could cause black women to receive lower wages than white women even if there were no racial differences in human capital investment patterns.(2) According to Becker, labor market discrimination against black women arises if employers have "tastes for discrimination"; that ls, they dislike hiring black women and only hire them if they can pay them less than white women. If discrimination is sufficiently widespread, then black women will receive lower wages than white women. The more intense and widespread employers' dislike of hiring black women, the lower black women's wages will be. This discriminatory behavior is costly to employers. An individual employer could hire only black women and lower his overall costs below what they would be if he were hiring a mix of black women and other workers. Thus, by practicing discrimination against black women, employers are foregoing potential profit opportunities. Since discriminatory behavior is costly to employers according to Becker's theory, it may not persist for long in highly competitive markets. Employers who do not discriminate will have higher profits than those who do and will have a tendency to expand and possibly to drive discriminatory employers out of business. In noncompetitive markets, however, employer discrimination can be a long-term phenomenon. Discrimination against black women can also arise when employers do not have tastes for discrimination, but black women's potential co-workers do. Co-workers have tastes for discrimination against black women if they demand a higher wage to work with black women than they would ask to work with other workers. When profit-maximizing employers are faced with a situation where discriminatory co-workers have the same skills as black women, they will react by separating black women from other workers (for instance, by placing them in different plants), but wage differences between black women and other workers will not necessarily arise. On the other hand, when the discriminatory co-workers have skills that complement the skills of black women (for instance, when the discriminatory co-worker is a floor manager and black women are applying for jobs as floor workers), segregation of the work force is not practical. Black women will be paid less than other workers with the same skills because they are more costly to hire: in order to hire black women, employers must raise the wages of the discriminatory co-workers above what they would have to pay them if they hired other workers in place of the black women. A similar result occurs when firms have discriminatory customers. Customers have tastes for discrimination if they are willing to pay less for goods and services produced by black women than for the identical goods and services produced by other workers. This type of discrimination is believed to be more frequent in the service sector than in other sectors of the economy because in the service sector customers come into direct contact with workers. When employers face customer discrimination, it is again more costly for them to hire black women than other workers, and as a result they pay black women lower wages. Unlike the case of employer tastes for discrimination, when the tastes for discrimination lie with co-workers or with customers, it is in the financial interest of employers to discriminate against black women by paying them less than they pay identical workers who are not black women. Moreover, there are no plausible mechanisms akin to competition for discriminatory firms for purging discriminatory workers or customers. It is, thus, possible for wage differentials due to discrimination from these sources to persist for a long time even in competitive markets. Even when it is in the employer's direct financial interest not to discriminate against black women, social and legal barriers can be erected to prevent employers from treating black women and other workers equally. A discriminatory society can socially ostracize nondiscriminatory employers or impose legal sanctions on them that are sufficiently high to alter their calculus and cause them to go along with discrimination against black women. Labor market discrimination by employers, co-workers, and customers can take other forms besides lowering black women's wages. Hiring discrimination by firms or managerial employees can make it difficult for black women to find employment, raising their unemployment rate in comparison to similarly qualified white women. If discriminatory employers, co-workers, or customers believe that some jobs or occupations are not appropriate for black women, black women may experience "occupational discrimination." Black women may be crowded into "acceptable" occupations or jobs. If this crowding results in an oversupply of workers to these occupations or jobs, wages in these occupations or jobs may become depressed, and black women will not only be in different jobs than their white counterparts, but they also will receive lower pay.(3) Alternative Theories of Discrimination: Statistical Discrimination and Efficiency Wages Alternative theories of discrimination point to possible reasons why black women might be paid less than other workers even if employers are strictly motivated by profit maximization, i.e., they do not have "tastes for discrimination" and are not prejudiced. The theory of "statistical discrimination" shows how differences in the average skill levels of black and white women can lead to differences in pay for black and white women with the same skills if employers have imperfect knowledge of their workers' individual skill levels.(4) Suppose that for some reason, unrelated to discrimination, black women on average invest less than white women in certain skills that employers find desirable, and also suppose that it is not possible for employers to ascertain an applicant's skill level at the time of making hiring decisions and wage offers. It is in the employers' financial interest to base their assessments of applicants on the available statistics and to treat each applicant as though she were average for her race. Employers will, thus, offer all black applicants lower wages than all white applicants (or alternatively refuse to hire them at all) based on their expectation that black applicants are likely to be less skilled than white applicants. Because pay is based on statistical average skill levels for each race, highly skilled black women will be offered lower wages than white women with low skills. This phenomenon is known as "statistical discrimination." In sum racial differences in human capital investment patterns can cause otherwise nondiscriminatory employers to discriminate by paying black women less than equally (and even less) skilled white women. Another theory thought to explain black-white wage differentials when employers are not prejudiced, the "efficiency wage theory," is usually developed in the context of a "dual labor market," i.e., under the assumption that jobs fall into two categories: primary sector jobs, in which workers' levels of effort are difficult to monitor, and secondary sector jobs, in which workers' levels of effort are easily discernible.(5) An example of a primary sector job might be a middle-level management position in a large company. An example of a secondary sector job might be working in a fast-food enterprise. Since employers find it difficult to monitor workers' effort in primary sector jobs, they cannot adjust individual workers' wages to their individual levels of effort. Instead, they adopt the following strategy to minimize shirking by their workers: they pay their workers more than for secondary sector jobs and fire any worker they discover shirking. This provides workers with an incentive not to shirk, because they do not want to be fired (if they were fired they would have to turn to the lower paying secondary sector for a job). In choosing the wage premium to pay their workers, primary sector employers choose an "efficient wage," or one that must balances out the extra wage costs from paying higher wages and the lower costs from reduced worker shirking. The efficiency wage theory might account for black women being paid less and working primarily in secondary sector jobs in the following way. Suppose that black women have higher job turnover rates than other workers. Since they are more likely to quit their job anyway than other workers, their benefit from a lifetime of higher wages in the primary sector is lower, and as a result to get them to shirk less employers would have to offer them even higher efficiency wages than for other workers. This makes it more costly for employers to hire black women than other workers for primary sector jobs and might explain why black women are only hired in the lower paying secondary sector. Interactions Between Discrimination and Human Capital Investment There are possible interactions between discrimination and human capital investment. The theories of statistical discrimination and efficiency wages show how racial differences in human capital investment patterns or in turnover rates can lead to discriminatory behavior on the part of unprejudiced employers. In turn discrimination in the labor market can cause black women to acquire less human capital than white women. It can affect their human capital investment directly by impeding black women's human capital investment opportunities or indirectly by lowering black women's returns to human capital investment. Discrimination in the educational system can directly limit black women's access to quality schooling, making it difficult or even impossible for black women to acquire an education comparable to white women's. Discrimination in the labor market can directly limit black women's access to on-the-job training, preventing black women from augmenting their skills. In a discriminatory environment, both informal training by co-workers and mentoring by superiors may be more difficult for black women to obtain. Moreover, wage discrimination of the type considered above can have an indirect, but important, negative effect on black women's human capital investment. If wage discrimination against black women is more severe at higher skill levels, then black women will have a lower return to human capital investment than other workers, and they will acquire less human capital. Similarly, if black women face occupational discrimination, they are unlikely to invest in skills useful only in jobs they cannot obtain. Thus, not only may the existence of discrimination cause black women to earn less than equally skilled white women, but it may also be responsible for black women being less skilled. Measuring the Effect of Discrimination on Black Women's Wages The discussion above indicates that the black-white female wage gap can be thought of as resulting from the joint effects of discrimination against black women and racial differences in women's human capital investment patterns. Moreover, racial differences in women's human capital investment patterns may themselves be partly caused by discrimination against black women. Thus, to measure the overall effect of racial discrimination on the black-white female wage gap, it is useful to distinguish between the direct effect of discrimination on wages and the indirect effect of discrimination on wages through its effect on skill formation. Part II attempts to measure the direct effect of labor market discrimination on black women's wages by measuring women's skills as carefully as possible and then comparing the wages of black and white women with the same skills. The direct effect of discrimination on black women's wages can be approximated by the extent that black women earn less than white women with the same skills. A limitation of this approach is that its accuracy depends on the extent to which women's skills can be measured. If important dimensions of skill are unmeasured, then this approach provides only an imperfect measure of the direct effects of labor market discrimination. Although the primary focus of part II is on measuring the direct effect of discrimination, part II also looks at discrimination's indirect effect by considering wherever possible whether discrimination is responsible for differences in the skill levels of black and white women. The indirect effect of discrimination, however, is much harder to quantify than the direct effect, and part II generally does not attempt to do so. ENDNOTES 1. This can happen if poor persons face borrowing constraints, termed "imperfect capita markets," that prevent them from borrowing to finance their education. It can also happen if wealthier families provide their children with more education for reasons unrelated to their future earnings (e.g., they want their children to be "well-rounded"). 2. Gary S. Becker, The Economics of Discrimination (Chicago: Univ. of Chicago Press, 1957) 3. See Barbara Bergmann, "Occupational Segregation, Wages and Profits When Employers Discriminate by Race or Sex," Eastern Economic Journal, vol. 1 (1974), pp. 103-10. 4. See E.S. Phelps, "The Statistical Theory of Racism and Sexism," American Economic Review, vol. 62 (1972), pp. 659-61; Kenneth J. Arrow, "Models of Job Discrimination," in Racial Discrimination in Economic Life, ed. Anthony H. Pascal (Lexington, Mass: Health & Co., 1972, pp. 83-102; Arrow, "Some Mathematical Models of Race in the Labor Market," in Racial Discrimination in Economic Life, pp. 187-204; Arrow, "The Theory of Discrimination," in Discrimination in Labor Markets, ed. Orley Ashenfelter and Albert Rees (Princeton, N.J.: Princeton Univ. Press, 1973), pp. 3-33; and D. J. Aigner and G.C. Cain, "Statistical Theories of Discrimination in the Labor Market," Industrial and Labor Relations Review, vol. 30 (1977), pp. 175-89. 5. See Jeremy I. Bulow and Lawrence H. Summers, "A Theory of Dual Labor Markets with Application to Industrial Policy, Discrimination, and Keynesian Unemployment," Journal of Labor Economics, vol. 4 (1986), pp. 376-414. Chapter 4 Determinants of Black-White Female Wage Differentials: Education and Geographic Location This chapter and chapter 5 use data from the U.S Censuses of Population to document a substantial convergence in the skills and characteristics of black and white women between 1940 and 1980. The censuses provide fairly consistent information on a wide array of characteristics for the entire U.S. population at 10-year intervals for the 1940-80 period. Unfortunately, they provide data on only some of the many dimensions of skill that are relevant to the labor market. The census data do allow documentation of trends in the education, occupations, regional distributions, and urban-rural mixes of black and white women over the 1940-80 period. In all of these factors, black women became more similar to white women between 1940 and 1980, and all of these factors are important determinants of wages. This chapter looks at the educational levels, regional distributions, and urban-rural mix of black and white women. Chapter 5 looks at their occupations. Education In 1940 the gap between the education of black women and the education of white women was enormous. On average, black women had completed 3 fewer years of school than white women. Almost two- thirds of black women had not completed the eighth grade, but more than three-quarters of white women had more than an eighth grade education. Only one-tenth of black women had graduated from high school, compared with one-third of white women. Tables 4.1 and 4.2 document trends in the years of school completed by black and white women between 1940 and 1980. Table 4.1 shows average educational levels by race for all women between the ages of 25 and 64 (by the age of 25 virtually all women have completed their schooling) and also for women in the 25-64 age group who were in the labor force. Table 4.2 shows the distribution of women by the number of years of schooling completed. Between 1940 and 1980, schooling levels increased substantially for both black and white women: the average years of school completed increased by over 3 years for white women and by over 5 years for black women. The proportion of women who had not finished eighth grade fell substantially for both groups. By 1980 over half of both black and white women had graduated from high school, and sizable proportions of both groups (15 percent of whites and 9 percent of blacks) had completed 4 years of college or more. Black women made even larger gains than white women over the period, and racial differences in schooling declined considerably. The 3-year difference between black and white women's average schooling levels in 1940 had shrunk to a difference of 1 year in 1980. The distributions of black and white women by years of school completed were markedly more similar in 1980 than they were in 1940. The timing of the racial convergence in education levels is also of interest. While increases in schooling levels progressed fairly smoothly over the entire period for both black and white women, black women gained very little relative to white women before 1950. Between 1940 and 1950, black women gained one-tenth of a year on white women. In each of the decades following 1950, black women gained at least half a year on white women. The larger relative increase in black women's schooling levels after 1950 was not due to a sudden improvement in black education at the time. Rather, it was caused by older black women who had been educated before 1920 progressively leaving the sample after 1950, and thus, it reflects much earlier increases in black schooling levels. In fact, table 4.3 shows that the relative schooling levels of black women between the ages of 25 and 34 began to increase in 1940, reflecting relative improvements in black schooling starting around 1920. The relative educational levels of black women in the 25-34 age group increased smoothly over the entire 1940-80 period. Black women in this age group gained about half a year of schooling on white women in every decade between 1940 and 1980. Because of the secular increase in women's educational levels. younger women, both black and white, were more educated than older women in each of the census years 1940-80, as can be seen in table 4.4. Moreover, because black women's educational levels were gaining on white women's, the difference in educational levels between black and white women was less at younger ages. In 1940 this difference ranged from 2.9 years for women aged 18-24 to 3.3 years for women aged 55 64. By 1980 the schooling gap for women aged 18-24 was .4, compared with a 1.9 year schooling gap for women aged 55- 64. Thus, the schooling levels for all black and white women converged considerably between 1940 and 1980, because less educated older black women progressively left the sample and the young black women entering the sample had progressively more education. The black- white convergence in schooling levels is even more pronounced when only women in the labor force are considered, as the lower panel of table 4.1 reveals. In 1940 black women in the labor force had an average of 3.6 fewer years of school than white women in the labor force. By 1980 this difference had narrowed to less than 1 year. Thus, in 1940 schooling differences between black and white women were greater for women in the labor force than for women in general. By 1980 schooling levels were closer for black and white women in the labor force than for women in general. Further, the convergence in schooling levels for black and white women in the labor force began by 1940, earlier than the convergence for women in general. The reason the schooling levels of women in the labor force are generally not the same as for the population as a whole is that labor force participation rates tend to differ by educational level. Table 4.5 shows labor force participation rates by educational level for black and white women between the ages of 25 and 64 for the 1940-80 period. Labor force participation rates of both black and white women increased with education throughout the period. There were, nevertheless, important differences between black and white women's labor force participation trends over the period. In 1940 black women with less than a 12th grade education were twice as likely as white women at the same educational level to participate in the labor force, and the difference between the labor force participation rates of black and white women declined with increased years of school. By 1980, on the other hand, the difference between the labor force participation rates of black and white women increased with years of school. Thus, the census data show an impressive convergence in the number of years of school completed by black and white women. Other researchers have found that the actual black-white convergence in schooling levels may have been even more pronounced than the census data suggest. Robert Margo and others have argued that the 1940 census overstates the actual time spent in school by older blacks.(1) Before 1920 the schools attended by blacks tended to be ungraded and for the most part held classes many fewer days per year than schools attended by whites. As a result, a black woman may have attended school for 8 years and only completed the equivalent of 5 years. The difference in the time spent in school by black and white women was, thus, probably substantially more than the 3 years reported in the 1940 census. Since later censuses more accurately reflected actual schooling levels of blacks, taking the 1940 census data at face value probably understates the degree to which black-white schooling levels actually converged over the 1940-80 period. Census data require measuring educational attainment by years of school completed (unadjusted for time spent in school per year) and provide no information on schooling quality, another important dimension of educational attainment. Using various other sources of information, other researchers have concluded that the average quality of education (including the length of the school year) received by blacks of working age probably increased both absolutely and relative to the quality of education received by whites over the 1940-80 period. Both the resources devoted to black education and the academic achievement (as measured by literacy rates, test scores, and so on) of black students appear to have improved steadily from the turn of the century at least into the 1960's. (2) Thus, the census data presented in this chapter do not capture the full measure of the increase in educational attainment of black women over the 1940-80 period. The true increase in black women's relative educational attainment was probably even more substantial than that suggested by the census figures on years of school completed. A preliminary look (3) at the relationship between wages and educational attainment over the period is shown in table 4.5, using census data on the years of school completed as the measure of educational attainment. Table 4.5 reports the average hourly wages for black and white women at different schooling levels for the years 1940-80. It also reports the ratio of black women's to white women's wages at each educational level. The narrowing of the black-white differential in the number of years of schooling completed over the 1940-80 period was undoubtedly responsible for part of the increase in black women's relative wages over the same period. The hourly wages received by both black and white women were higher the more years of schooling completed. (4) Moreover, the ratio of black women's wages to white women's wages also increased with the number of years of school completed: in 1940, for instance, black women with less than a 12th grade education earned only half as much as white women with the same amount of schooling. but college-educated black women earned 80 percent as much as college-educated white women. Thus, black women's wages probably would have increased relative to white women's over the 1940-80 period due to their increased relative schooling amounts, even if there had been no other changes over the same period. It is also apparent from the data in table 4.6 that the convergence in the number of years of school of black and white women cannot by itself explain the entire increase in black women's relative wages over the period. Over the years 1940-80 black women's wages increased relative to white women's at all three educational levels. The relative wage of black women with less than a 12th grade education increased from 48 percent to approximately 90 percent; the relative wage of black women with a high school education increased from 60 percent to more than 95 percent; and the relative wage for college-educated black women increased from 80 percent to more than 100 percent. Thus, the overall relative wage of black women would have increased over the 1940-80, period even if their relative educational attainment had not increased at all. One possible explanation for the increased relative wages of black women within schooling levels is that the quality of schooling increased at each level over the period. It is likely that other factors, such as a decline in discrimination and increases in other skills possessed by black women, also played a role. In sum, the 40-year period between 1940 and 1980 witnessed an impressive increase in the relative educational attainment of black women. Although the census data only allow documentation of increases in relative schooling levels, the relative quality of black women's education probably improved as well. The increased educational attainment of black women was probably responsible for some but not all of the decrease in the black-white female wage gap over the period. Region and Urban or Rural Location Black women have always been much more likely to live in the South (5) than white women. Since wages have historically been lower in the South than elsewhere, the different regional distributions of black and white women are one factor that has contributed to black women's lower relative earnings. Similarly, in the past, black women were more likely than white women to live in low-wage rural areas, another factor contributing to their lower relative earnings. This section examines the changing regional distributions and urban-rural mixes of black and white women over the years 1940-80. Table 4.7 shows that there was a substantial convergence in the regional distributions and a major shift in the urban-rural mixes of black and white women from 1940 to 1980. In 1940 almost three-quarters of black women and only one-quarter of white women lived in the South. Over the following years, especially between 1940 and 1960, there was a large northward migration of black women, so that by 1980 the proportion of black women living in the South had declined to just over one-half. At the same time, there was a much smaller white migration to the South, so that by 1980 almost one-third of white women lived in the South. Thus, the geographic distributions of black and white women were much more similar in 1980 than they had been in 1940. In 1940 black women were somewhat more likely than white women to live in rural areas. Over the years following, the proportion of black women living in rural areas decreased considerably, from more than 50 percent to less than 20 percent, compared with a much smaller decline (from 40 to 30 percent) in the proportion of white women living in rural areas. Thus, by 1980 black women were less likely than white women to live in rural areas. The shift in urban-rural patterns by race was partially due to the northward migration of blacks over the 1940-80 period and partially due to a movement by black women from rural to urban areas within the South. Throughout the period, black women living outside of the South were extremely unlikely to live in rural areas: between 90 and 97 percent of black women living outside of the South lived in urban areas. By contrast about two-thirds of southern black women in 1940 lived in rural areas. Thus, for many black women, northward migration meant a rural to urban move as well. At the same time, the proportion of southern women, both black and white, who lived in rural areas was declining. About two-thirds of southern women of both races lived ln rural areas in 1940. By 1980 the fraction of southern women living in rural areas had shrunk to less than one-third for black women and to 40 percent for white women. Thus, not only did black women move from the predominately rural South to the heavily urban (at least for blacks) non-South over the 1940-80 period (compared with a small reverse migration among white women), but at the same time, the black women who remained in the South moved to urban areas to a greater extent than white women. Table 4.8 presents estimates derived from census data of the average hourly wages of black and white women by region for the years 1940-80. Not only did wages outside of the South exceed those in the South for both races throughout the period, but black women earned more relative to white women outside of the South than in the South. Thus, the northward migration of black women combined with the more stable regional distribution of white women over the years 1940-80 may have contributed to black women's increased relative earnings over this period. On the other hand, black-white wage ratios increased within both regions of the country (more in the South than in the non-South), indicating that changes in living patterns were not solely responsible for black women's increased relative wages over the period. Table 4.9 presents the hourly wages of black and white women by whether or not they lived in rural areas. As expected, the hourly wages of both black and white women living in rural areas were lower than those of their counterparts living in urban areas. Moreover, black women's wages were lower relative to white women's wages in rural areas than in urban areas. Thus, black women's shift from being a more rural population than white women in 1940 to being a more urban population than white women in 1980 may also have contributed to their increased relative wages over these years. Since black women living in rural areas also lived predominately in the South (table 4.7), it is possible that black women's relatively low wages in rural areas were simply a reflection of the southern residence of black women living in rural areas. Thus, the figures presented in tables 4.8 and 4.9 do not allow full identification of the separate effects of region and urban or rural residence on black women's relative wages. Similarly, the figures in table 4.10 point to a problem in using the evidence presented so far to distinguish the separate effects on wages of region and education. Table 4.10 presents the average education of black and white women living in the South and the non-South for the years 1940-80. Women in the South were on average less educated than their counterparts elsewhere. Moreover, southern black women were less educated relative to their white counterparts than were black women living elsewhere. In 1940 southern black women had completed 3 fewer years of school than southern white women; the black-white schooling gap outside of the South was only 1 1/2 years. The black-white schooling gap decreased in both regions (more in the South than in the non-South) over the 1940-80 period. In 1980 southern black women still had completed 1 year less of school than their white counterparts, compared with seven tenths of a year less of school for black women living outside of the South. The schooling trends observed in table 4.10 are one possible explanation for the regional patterns in the wages of black and white women noted in table 4.8. Women's wages in the South may have been lower simply because southern educational levels were lower than in the rest of the country. Black women may have had lower relative wages in the South simply because the black-white schooling gap was larger in the South. The increasing black-white wage ratio within the South and the increasing earnings for black women in the South relative to the non-South may have been caused by the increasing southern black schooling levels over the years 1940-80. Thus, although the figures presented in table 4.8 suggest that region played an important role in determining racial wage patterns for women over the years 1940-80, they do not disentangle the separate effects of education and region. Summary Schooling and geographical location (region and urban-rural residence) patterns of black and white women are potential explanations for black-white female wage trends over the years 1940-80. The relatively low educational levels of black women and their heavy representation in low-wage southern and rural areas could each be partially responsible for the relatively low wages earned by black women. The convergence in black-white schooling levels, regional distributions, and urban-rural mixes over the years 1940-80 could each be partially responsible for black women's increasing relative wages over the period. The next chapter considers another possible determinant of women's wages: their occupations. ENDNOTES 1. See Robert A. Margo, "Race, Educational Attainment, and the 1940 Census," Journal of Economic History, vol. 46 (1986), pp. 189-98, and "Race and School Attendance in the American South: Evidence from the 1900 Census Sample" (Mimeo. Philadelphia: Univ. of Pennsylvania, 1986); and James P. Smith, "Race and Human Capital," American Economic Review, vol. 77 (1974), pp. 685-98. 2. See U.S Commission on Civil Rights, The Economic Progress of Black Men in America (1986), pp. 54-72 for a thorough discussion of racial differences in schooling quantity and quality. 3. See chap. 6 for a more thorough evaluation of the impact of the increased educational attainment of black women on the black-white female wage gap. 4. Freeman has also found that black women's earnings increase relative to white women's as education rises. Richard B. Freeman, "Decline of Labor Market Discrimination and Economic Analysis," American Economic Review, vol. 63, no. 2 (1973), pp. 280-86; and "labor Market Discrimination: Analysis, Findings and Problems," in Frontiers of Quantitative Economics, ed. Michael D. Intriligator and David A. Kendrick (Amsterdam: North-Holland, 1974), chap. 9. 5. The South is defined in this report as the States of Delaware, Maryland, Virginia, North Carolina, South Carolina, Georgia, Florida, West Virginia, Kentucky, Tennessee, Alabama, Mississippi, Louisiana, Arkansas, Texas, and Oklahoma and the District of Columbia. Chapter 5 Determinants of Black-White Female Wage Differentials: Occupations There are two kinds of females in this country--colored women and white ladies. Colored women are maids, cooks, taxi drivers, crossing guards, schoolteachers, welfare recipients, bar maids and the only time they become ladies is when they are cleaning ladies. Louise D. Stone, "What It's Like to Be a Colored Woman," Washington Post, November 13, 1966 Throughout history, black and white women have worked in different jobs. Statistical analysis of census data cannot provide a full measure of the extent of work force segregation by race, nor can it fully capture the differences in working conditions and employer treatment experienced by black and white women. Nevertheless, the study of racial differences in occupations over the years 1940 80 in this chapter provides a more complete picture of racial differences in economic status than would a study focusing on wages alone. The analysis of racial differences in occupations is particularly important because occupations are often a basis for discrimination. The convergence in black and white women's occupations between 1940 and 1980 was perhaps even more striking than the convergence in their levels of educational attainment and in their regional distributions. In 1940 black women were confined to a small number of low-status occupations. Seventy percent of black women were employed in just two occupations: domestic servant and farm laborer. Fewer than 10 percent of black women worked in middle- or high-status occupations. By contrast, white women worked in a variety of occupations, and more than one-half of white women worked in middle- and high-status occupations. By 1980 black women had made mayor inroads into occupations, such as clerical work, in which they were virtually unrepresented in 1940. However, important differences in the occupational distributions of black and white women remained, and black women continued to be overrepresented in low-status occupations and underrepresented in middle- and high-status occupations. Occupational Distributions of Black and White Women: 1940-80 Table 5.1 shows the occupational distributions of black and white women for the years 1940, 1960, and 1980. (1) In 1940 black women's and white women's occupations were almost completely distinct. Black women were almost exclusively employed in low- status occupations, whereas white women were largely employed in middle- and high-status occupations. Almost 60 percent of black women worked as domestic servants (compared with 8 percent of white women). Another 11 percent of black women were farm laborers (compared with less than 1 percent of white women). Two other occupational categories that employed sizable percentages of black women were service occupations (about 10 percent) and operatives (8 percent), but both of these categories employed larger percentages of white than of black women. Black women were virtually unrepresented in the middle-status clerical and sales occupations, which together employed almost one-third of white women. Clerical work, employing almost one-quarter of white women, was the largest single occupational category for white women but employed less than 2 percent of black women. Eight percent of white women were employed in sales occupations, compared with less than 1 percent of black women. Only 1 in 20 black women was employed in the high-status professional, technical, and managerial occupational categories, compared with 1 in 4 white women. Fewer than 5 percent of black women, but almost 20 percent of white women, were in professional and technical occupations. Furthermore, whereas white women were spread across a number of professional occupations, black women professionals were almost exclusively teachers. (Four-fifths of black women professionals were teachers, compared with one-half of white women.) Moreover, although nursing was a sizable profession for white women, there were almost no black nurses in 1940. Similarly, although almost 5 percent of white women were employed in managerial occupations, there were very few black women managers. Some changes in black women's occupations were visible by 1960, but the large majority (80 percent) of black women continued to be employed in low status occupations. The proportion of black women working as domestic servants declined by 20 percentage points between 1940 and 1960, to 38 percent. The proportion of black women employed as farm laborers also declined considerably, to just over 3 percent. Over the same years, the proportion of black women employed as service workers more than doubled, from 10 to 24 percent, and the proportion employed as operatives almost doubled, from 8 to 15 percent. These changes reflected the large northward migration of blacks and a decline in the demand for domestic servants after World War II. They also represented a small improvement in the occupational status of black women. (2) At the same time, black women were beginning to make small inroads into a few middle- and high-status occupations. The proportion of black women doing clerical work increased from 1.3 percent to 8.0 percent over the 20- year period between 1940 and 1960. However, the proportion of black women who were clerical workers remained well below that of white women, which increased from one- quarter to one-third over the same 20-year period. The proportion of black women employed as professional and technical workers increased by two-thirds, from 4.6 percent to 7.7 percent. Although much of this increase occurred among teachers, the proportions of black women in nursing and in other professional and technical occupations increased noticeably (but remained small). In addition, black women continued to be unrepresented in both managerial and sales occupations. The slight gains in occupational status made by black women between 1940 and 1960 gave way to major improvements between 1960 and 1980. The proportion of black women working as domestic servants decreased dramatically from 38 to 6 percent. Unlike the 1940 to 1960 period, this decline was not accompanied by an increase in the proportions of black women employed in other low- status occupations, such as service workers and operatives. Instead, the proportions of black women employed in middle- and high- status occupations increased considerably: the proportion of black women employed as clerical workers increased almost four-fold to 29 percent, only 7 percentage points below the comparable figure for white women. The proportion of black women employed in professional and technical occupations doubled from about 8 percent to 16 percent, only 4 percentage points below the comparable figure for white women. The movement of black women into professions other than teaching, first apparent in 1960, continued between 1960 and 1980. Almost 3 percent of black women were nurses in 1980 and more than 5 percent were in other professional and technical occupations. Black women made major occupational gains over the 1940 to 1980 period, but differences remained between the occupations held by black and white women in 1980. Black women continued to be less well represented in high status occupations than white women. Black women were less likely to be professionals than white women, and female black professionals were slightly more likely to be teachers than female white professionals. Black women were less than half as likely as white women to be in managerial occupations. Black women also continued to be under-represented in middle-status occupational categories. Black women were less likely to be clerical workers than white women. In particular, they were much less likely to be in secretarial jobs. Although some black women had succeeded in entering sales occupations, black women were less than half as likely to be sales workers as white women. Black women in 1980 also continued to be more likely than white women to be employed in low-status occupational categories, especially as operatives, service workers, and domestic servants. One way to assess the extent of the differences in the occupational distributions of black and white women in a given year is to calculate the "index of occupational dissimilarity." which represents the percentage of black women (or white women) who would have to change occupations in order for black and white women's occupational distributions to be the same. (3) Values of the index that are closer to zero, thus, correspond to more equal occupational distributions for the two groups and values of the index that are closer to 100 correspond to more unequal occupational distributions. Table 5.2 reports the values of the index of occupational dissimilarity for black and white women for the years 1940, 1960, and 1980. To provide benchmarks for comparison, values of the index for white women and white men, for black women and black men, and for black men and white men are also shown. The 1940 value of the index of occupational dissimilarity for black and white women was 63.1, indicating that almost two-thirds of black (or white) women would have had to switch occupations in 1940 to equalize the occupational distributions of the two groups. The occupational distributions of black and white women were even more dissimilar than were the occupational distributions of white men and white women. They were also considerably more dissimilar than the occupational distributions of white men and black men. Between 1940 and 1960, the index of occupational dissimilarity for black and white women declined only slightly, but between 1960 and 1980, there was a large decrease in the index. By 1980 the occupational distributions of black and white women had become much more similar than the occupational distributions of white men and white women. Moreover, unlike in 1940, the occupational distributions of black women and white women were closer than those of black men and white men. Black women's relative wage over the 1940-80 period was closely tied to their occupational status. Chapter 6 shows that the different occupational distributions of black and white women account for much of the black-white female wage gap during this period and that black women's improving occupational status- between 1940 and 1980 was responsible for much of their increased relative wages during these years. Thus, in order to evaluate the effect of discrimination on the economic status of black women, it is important to consider the sources of black-white occupational differences. Were black women in low-status occupations because they were un-skilled, uneducated, and lived in the rural South, or were they in low-status occupations because they were barred by discriminatory employment practices from working in middle-and high-status occupations. (7) The remainder of this chapter addresses this important issue. Occupational Distributions of Black and White Women by Education: 1940-80 Since higher status occupations tend to require more education. the large differences in the educational attainment of black and white women noted earlier might account for the differences in their occupations. Tables 5.3, 5.4, and 5.5 show the occupational distributions of black and white women by educational level for 1940, 1960, and 1980, and table 5.6 shows the values of the index of occupational dissimilarity for black and white women by educational level for the same years. These tables reveal that even within educational levels black and white women had very different occupational distributions in all 3 years. Thus, black women's lower occupational status was not simply a reflection of their lower educational attainment. The difference between the occupations of black and white women was greater for less educated women in all 3 years. The occupational distributions of black and white women within educational levels became more similar over time at all education levels, and this happened earlier for highly educated black women than for less educated black women. There were striking differences in the occupations of black and white women at all educational levels in 1940 (see table 5.3). Although some of these differences might reflect the lower quality of the education received by blacks or artificially high black schooling levels reported in the 1940 census. the differences are far too large to be fully accounted for by these factors. Black women were not domestic servants in 1940 simply because they were uneducated: almost one-quarter of black women with some post-high school education and even 13 percent of black women with a college degree were employed as domestic servants. Similarly, the low educational attainment of black women was not the reason for their almost total absence from the field of clerical work. Whereas 40 percent of white women with a 12th grade education and 30 percent of white women with some post-high school education were employed as clerical workers, the comparable figures for black women were 7 and 6 percent, respectively. Moreover, white women with low levels of education found employment in occupations. such as operatives and service work, in which black women, regardless of education, were unlikely to be employed. Highly educated black women appear to have found some opportunities that were unavailable to black women with less education. Indeed, the value of the index of occupational dissimilarity for black and white women with a college degree was 28, and the value for women with some post-high school education was 48, compared with values of the index between 55 and 60 at lower educational levels (see table 5.6), and black women with more than a high school education were just as likely as white women to be employed in professional and technical occupations. However, within the professional and technical occupational category, black women were much more likely than white women to be employed as teachers and less likely to be employed in any other occupation. Moreover, If highly educated black women did not find employment as teachers, they were more likely to be employed as domestic servants than as clerical workers, the second largest occupational category for white women with more than a high school education. Thus, the greater degree of occupational equality experienced by highly educated black women was probably due to the demand for black school teachers generated by segregated school systems. Like their less educated counterparts, highly educated black women in 1940 were excluded from many occupations open to white women with the same amount of education. From 1940 to 1960, the differences between the occupational distributions of black and white women narrowed considerably within higher educational levels but barely changed within lower educational levels (see table 5.6). New employment opportunities opened up for highly educated black women during this period. For the first time, large proportions of black women with a high school education or more (almost 20 percent of black women with 12 years of school and more than 30 percent of black women with 13 to 15 years of school) found jobs in the clerical sector. Professional occupations such as nursing and library, social, and religious work, among others, also began to employ noticeable proportions of educated black women. On the other hand, there was a large decline in the proportion of black women with 13 to 15 years of education who were teachers, probably because of increased standards for teachers in black schools, and this decline was not completely offset by increases in the proportion who were in other professional occupations or in the clerical sector. More-over, black women with a high school education or more continued to be much more likely than white women to work in the service sector or as domestic servants. Black women with less than a 12th grade education found many fewer new opportunities in the 1940-60 period. Decreases in the proportions of these women employed as farm laborers and as domestic servants were largely offset by increases in the proportion employed as service workers. Although the proportion of these women hired as operatives increased some, black women with low levels of education remained much less likely than comparable white women to be employed as operatives. Between 1960 and 1980, the occupational distributions of black and white women became more similar at all educational levels (see table 5.6). In contrast to the changes that occurred between 1940 and 1960, the changes between 1960 and 1980 were greater for less educated women, largely because less educated black women left jobs as domestic servants for jobs held by white women, such as service work or jobs as operatives, and for high school graduates, clerical work. Despite these changes, the occupational distributions of black and white women remained more dissimilar at lower educational levels than at higher educational levels in 1980. A possible reason for the greater occupational dissimilarity among the less educated is that on average these women are older (4) and may have made their occupational choices prior to the opening up of new job opportunities for black women. Although black women's lower educational attainment in 1940 would undoubtedly have limited their occupational opportunities somewhat in any case, the extreme dissimilarity of black and white women's occupations at that time implies that other factors, such as discrimination against black women, played a far greater role than racial differences in educational attainment in keeping black women out of occupations commonly held by white women. In fact, occupational simulations show that even if black women had been as educated as white women in 1940, the occupations held by black and white women would have been almost as different as they actually were. (5) Indeed, if black women in 1940 had had white women's educational attainment instead of their own, the percentage of black women who were domestic servants would have been reduced by only 10 percentage points to 50 percent (8 percent of white women were domestic servants in 1940). Factors other than education continued to be important in limiting black women's occupational opportunities in 1960, especially for less educated black women. However, the increasing occupational similarity of black and white women within education groups between 1960 and 1980 suggests that the role of other factors, such as discrimination, diminished considerably after 1960. Nonetheless, there continued to be noticeable occupational differences within education groups between black and white women in 1980, particularly among less educated women. This suggests that race-based occupational discrimination against black women may not have entirely disappeared by 1980, although to the extent that the less educated are also older, occupational dissimilarities for them may be due instead to past labor market discrimination. Occupational Distributions of Black and White Women by Region: 1940-80 To determine the extent to which the large black-white differences in occupational distribution over the 1940-80 period resulted from occupational discrimination against black women, it is necessary to explore the possibility that black-white differences in characteristics other than education may have been partially responsible. As noted above, there were important differences in the regional distributions of black and white women during the years 1940-80, particularly at the beginning of the period. If regional economies differed, causing job opportunities for women to be different in the South than in the rest of the country, then the high proportion of black women living in the South could account for some of the observed occupational differences between black and white women. Tables 5.7. 5.8, and 5.9 show the occupational distributions of black and white women by region (South and non-South) for 1940, 1960, and 1980, and table 5.10 shows the values of the index of occupational dissimilarity for black and white women by region for the same years. The figures in these tables indicate that black women's limited occupational opportunities over the 1940-80 period cannot be explained by their predominately southern residence. White women had virtually the same occupational distributions whether they lived in the South or not, implying that the distribution of Job opportunities for women as a group was not markedly different in the South than in the rest of the country. Moreover, the black-white differences in occupational distributions within regions were comparable in magnitude to the overall differences in black-white occupational distributions, indicating that the regional distributions of black and white women account for little of their occupational differences over the 1940-80 period. The indices of occupational dissimilarity shown in table 5.10 were greater in the South than in the non-South in all 3 years, suggesting that black women generally had better occupational opportunities relative to white women outside of the South. Black-white occupational differences decreased over time both in the South and in the non-South, but these decreases happened earlier and were larger outside of the South. There was a noticeable convergence in the occupations held by black and white women outside of the South between 1940 and 1960, and almost no convergence in the South. Both regions experienced substantial black-white convergence in occupations between 1960 and 1980, but black-white occupational differences remained greater in the South than in the rest of the country in 1980. Black women in the South faced greater occupational barriers than black women in the non-South in 1940. In both regions of the country about 60 percent of black women (and fewer than 10 percent of white women) were domestic servants (see table 5.7). In the non-South, black women who were not domestic servants found employment as operatives and service workers. In the South, on the other hand, black women found many fewer opportunities in these occupations and instead became farm laborers. By contrast, white women were equally likely to be operatives and service workers in both regions, and almost no white women worked as farm laborers in either the South or the non-South. The textile industry provides a particularly compelling example of the degree to which black women in the South faced more limited job opportunities than black women in the rest of the country. Fewer than 1 percent of southern black women worked as textile operatives, compared with 13 percent of southern white women. Whereas white women were almost twice as likely to be employed as textile operatives in the South as in the non-South, black women were 6 times as likely to be employed as textile operatives in the non-South as in the South. These figures reflect a pattern of extremely strong discrimination against black women in the southern textile industry. Jim Crow laws formalized the discrimination against black women in the southern textile industry. A South Carolina law passed in 1915 and still in effect in the 1960s forbade employers to allow black and white employees to "labor and work together within the same room or to use the same doors of entrance and exit at the same time, . . . or to use at any time the same lavatories, toilets, drinking water buckets, pails, cups, dippers or glasses...." Laws such as this made it very costly for employers to hire black and white workers in the same capacity. Even absent Jim Crow laws, employers were under strong pressure not to hire black workers. If employers were to hire a black worker, their white workers would quit en masse, and "it became almost impossible to get whites to use a mill house formerly occupied by the Negro." (6) Teaching was the one exception to the rule that black women had fewer occupational opportunities in the South than the rest of the country in 1940. Black women were almost twice as likely to be teachers in the South than in the non-South, whereas white women were about equally likely to be teachers in the two regions. Throughout the South there were usually two separate school systems, one white and one black. Either by custom or by law, most of the school teachers in black schools were blacks. (7) As a result there was greater demand for black school teachers in the South than in the rest of the country. Between 1940 and 1960, more new opportunities opened up for black women outside of the South than in the South. Most notably, the field of clerical work began to employ a sizable proportion (13.7 percent) of black women outside of the South, but continued to employ very few southern black women (3.8 percent). Black women in the non-South also substantially increased their employment as operatives (from 14.5 percent in 1940 to 21.5 percent in 1960) and as service workers (from 14.3 percent in 1940 to 25.2 percent in 1960). The new job opportunities outside of the South were accompanied by a large drop in the proportion of black women employed in very low-status occupations. The proportion of black women outside of the South employed as domestic servants dropped from 59.4 percent in 1940 to 26.0 percent in 1960. In the South the proportion of black women employed in very low-status occupations also dropped between 1940 and 1960, but to a much lesser extent. The proportion of southern black women who were domestic servants dropped from 58.1 percent to 46.4 percent, and the proportion of black women employed as farm laborers dropped from 15.1 percent to 5.1 percent. These changes were accompanied by a large increase in the proportion of southern black women who were service workers (from 8.7 percent in 1940 to 22.4 percent in 1960) but by only small increases in the proportions who were operatives (from 5.7 percent to 9.6 percent) and clerical workers (from 0.7 percent to 3.8 percent). As in 1940, the southern textile industry employed extremely few black women. The occupational barriers faced by black women were further reduced between 1960 and 1980 in both regions of the country. However, despite significant improvements in black women's opportunities in the South (for instance, the southern textile industry, which had employed virtually no black women in 1940 and 1960, employed a larger percentage of black women than of white women in 1980), black women in the South continued to have relatively more restricted opportunities than black women in the rest of the country.8 Most notably, black women in the South continued to be underrepresented in the clerical sector. Whereas approximately equal proportions of white women were clerical workers in the South and in the non-South (37 percent in the South and 36 percent in the non-South), black women were much less likely to be clerical workers in the South than in the rest of the country (23 percent in the South and 35 percent in the non-South). The greater occupational dissimilarity of black and white women in the South than the non-South during the years 1940-80 indicates that discrimination against black women was greater in the South than in the rest of the country. On the other hand, at least some of the regional differences in black-white occupational dissimilarity might also be accounted for by factors other than occupational discrimination. Black women in the South lived predominately in rural areas, and black women outside of the South lived almost exclusively in urban areas, whereas the urban-rural mix for white women was fairly similar across regions. Thus, the regional differences in relative job opportunities for black and white women found above could be the result of differences in the types of jobs available in urban and rural areas. Similarly, since black women had lower educational attainment in the South than in the non- South, and white women had fairly equal educational attainment across regions, differences in the types of jobs open to women at different educational levels could show up as regional differences in job opportunities. However, even if black women had the same educational attainment and urban-rural mix as white women within each region, large regional differences in the relative occupational opportunities of black women would have existed throughout the 1940-80 period. (9) Another possible explanation for the greater occupational dissimilarity of black and white women in the South is that the education received by black women may have been of lower quality relative to the education received by white women in the South than in the rest of the country. Unfortunately. the quality of education cannot be measured from census data. Other data sources do provide some evidence about the quality of the education received by black women in the South, at least for the contemporary period. In a nationally representative sample, black women with 12 years of education and above were found to have lower scores on achievement tests relative to white women with comparable educational attainment in the South than in the rest of the country.(10) This suggests that at higher educational levels (12 years of education and above) the relative quality of the education received by black women in the South may, indeed, be worse than in the rest of the country. This might account for part of southern black women's relatively low representation in the clerical sector, since clerical workers generally have relatively high educational levels. On the other hand, the relative test performance of black women who are less educated appears to be no worse in the South than in the rest of the country. (11) Consequently, regional differences in the relative quality of the education received by black women are unlikely to account for their historically lower representation among operatives in the South, since operatives generally have relatively low levels of education. Thus, the relatively more limited occupational opportunities of southern black women in 1980 might be caused by the relatively low quality of black education in the South. It is unlikely, however, that the relatively low quality of southern black education is primarily responsible for the limited opportunities of southern black women in 1940 and 1960, because in these years black women were excluded even from occupations requiring very little education. Greater discrimination in the South than in the rest of the country was probably the most important factor limiting southern black women's relative occupational opportunities in those years. In sum, differences in the regional distributions of black and white women cannot account for the differences in their occupations over the years 1940-80, because there were large black-white occupational differences within regions during this period. Black- white occupational differences declined over the period both in the South and in the rest of the country, but black-white occupational differences were larger in the South throughout the period. The greater black-white occupational differences in the South cannot be accounted for by regional differences in measurable characteristics, such as the educational attainment and urban-rural mix, of black and white women. In 1940 and 1960, greater black-white occupational differences in the South were probably primarily due to greater occupational discrimination against black women in the South than in the rest of the country. By 1980 regional differences in occupational discrimination appear to have diminished, and greater black-white occupational differences in the South may have been due primarily to the lower relative quality of education received by black women in the South. Occupational Distributions of Black and White Women Controlling for Characteristics: 1940-80 To assess the combined effect of racial differences in characteristics such as educational attainment, regional distribution, urban-rural residence, and age (all characteristics readily available from census data) on black-white occupational differences, this section asks the following questions: What would black women's occupational distributions in the years 1940-80 have looked like if they had had white women's characteristics? What would white women's occupational distributions have looked like if they had had black women's characteristics? Hypothetical occupational distributions were generated for black and white women assuming that each group had the other group's characteristics. Comparisons of these hypothetical occupational distributions with the actual occupational distributions of black and white women help clarify whether black women's lower occupational status was due to racial differences in characteristics or instead to discrimination limiting black women's access to occupations. Tables 5.11-5.13 show hypothetical distributions for black and white women assuming that they had the other group's characteristics for the years 1940, 1960, and 1980. For comparison, black and white women's actual occupational distributions for these years are also shown. Table 5.14 shows values of the index of occupational dissimilarity for various comparisons of the hypothetical occupational distributions with the actual occupational distributions of black and white women. These tables show that differences in the characteristics of black and white women account for only a small portion of the black-white occupational differences over the years 1940-80. In 1940, for instance, even if black women had had white women's educational attainment, regional distribution, urban-rural mix, and age distribution, the index of occupational dissimilarity would have been reduced only from 64.0 to 47.7. More than 50 percent of black women would have been domestic servants. (13) Giving black women white women's characteristics would have increased their employment as textile operatives from 2.3 to 4.5 percent (primarily because a higher proportion of white women lived outside of the South and black women were more likely to be hired as textile operatives outside of the South), but giving white women black women's characteristics would have almost tripled their employment as textile operatives, from 8.5 to 24.3 percent. These figures appear to confirm that the textile industry discriminated against black women, particularly in the South. In 1980 black-white differences in characteristics continued to explain only a small portion of their occupational differences, primarily because black women's and white women's educational attainment, regional distributions, urban-rural mix, and age distribution were much closer in 1980 than they had been in 1940. If black women had had white women's characteristics, their occupational distribution would have been fairly close to that of white women. However, black women would still have been underrepresented among managers (2.7 percent of black women would have been managers if black women had white women's characteristics, compared with 6.4 percent of white women); in the clerical sector (30.1 percent of black women would have been clerical workers, compared with 36. 5 percent of white women), and among sales workers (3.0 percent of black women would have been sales workers, compared with 6.9 percent of white women). Black women would have been overrepresented among operatives and service workers (10.6 percent of black women would have been operatives compared with 7.2 percent of white women, and 22.5 percent of black women would have been service workers compared with 15.4 percent of white women). In sum, very little of the occupational differences between black and white women over the years 1940-80 can be accounted for by racial differences in the characteristics that can be measured using census data (educational attainment, regional distribution, urban-rural mix, and age distribution). Even if black women had white women's characteristics, they would have been more likely to work in low-status occupations (as domestic servants early in the period, then as service workers and operatives later on) and less likely to work in middle- and high-status occupations (particularly clerical occupations) than white women. What do these results imply about the degree to which occupational discrimination against black women limited their occupational opportunities during the years 1940-80? In the early years of the period (1940-60), the differences between the hypothetical occupational distributions for black women assuming that they had white women's characteristics and white women's actual occupational distributions were so large that it is clear that occupational discrimination had a sizable effect on black women's occupations. There may well have been black-white differences in characteristics that were omitted from the analysis because they could not be measured using census data. For instance, the lower quality of education received by black women might account for some of the black-white occupational gap. Another omitted characteristic, previous work experience, however, was probably greater for black women than for white women and works to narrow the gap. Thus, it is extremely unlikely that differences in omitted characteristics, however substantial, could fully account for differences as large as those that existed between the hypothetical black occupational distributions and the actual white occupational distributions in the years 1940 and 1960. By 1980, however, the differences between the hypothetical black occupational distribution and white women's actual occupational distribution had narrowed considerably. Thus, in 1980 it is possible that racial differences in unmeasured characteristics, such as the quality of education, might account for the remaining black-white occupational gap. On the other hand. it is also possible that some residual occupational discrimination against black women continued to exist even in 1980. (14) Changes In Black Women's Occupational Distribution 1940-80: Discrimination or Characteristics? As shown above, black women's occupations changed considerably between 1940 and 1980. (15) To determine whether black women's improving occupational status was due to changes in their characteristics--increased educational attainment, changing geographical distribution, movement away from rural areas, and so on--or rather to a decrease in the extent of occupational discrimination against black women, hypothetical occupational distributions were generated to answer the following questions: What would the occupational distribution of black women in 1940 have looked like if they had had the characteristics of black women in 1980? What would the occupational distribution of black women in 1980 have looked like if they had had the characteristics of black women in 1940. Table 5.15 shows these hypothetical occupational distributions and, for comparison, black women's actual occupational distributions in 1940 and 1980. Table 5.16 shows values of the index of occupational dissimilarity for comparisons of the hypothetical occupational distributions and black women's actual occupational distributions in 1940 and 1980. Tables 5.15 and 5.16 show that black women's changing characteristics were responsible for only a small portion of their improved occupational status over the period 1940-80. For instance, giving black women in 1940 the characteristics of black women in 1980 would have decreased the proportion of black women working as domestic servants only slightly, from 58.4 percent to 45.1 percent, whereas the actual 1980 proportion was 6.2 percent. Similarly, it would have increased the proportion of black women working in the clerical sector only slightly, from 1.3 percent to 5.6 percent, whereas the actual 1980 proportion was 29.0 percent. If black women in 1940 had had the same characteristics as black women in 1980, their overall occupational distribution would have changed slightly, but the difference would have been nowhere near as large as the actual difference between black women's 1940 and 1980 occupations. (16) Thus, black women would have experienced only a slight improvement in occupational status in 1940 if the only thing that changed between 1940 and 1980 had been black women's characteristics. The changing characteristics of black women can account for only a small part of the improved occupational status experienced by black women between 1940 and 1980. Declining occupational discrimination against black women was, thus, probably responsible for most of their improved occupational status over the period. Summary Black women's and white women's occupations converged substantially over the years 1940-80. At the beginning of the period, black women were confined to extremely low-status occupations. By the end of the period, black women were better represented in middle- and high-status occupations, and their overall occupational status was only slightly lower than white women's. Although racial differences in educational attainment, regional distribution, urban-rural mix, and age can partially account for black women's lower occupational status over the period, it is likely that occupational discrimination against black women played a far greater role in limiting black women's access to occupations commonly held by white women, particularly in 1940 and 1960. Indications were found that occupational discrimination against black women was more severe in the South than in the rest of the country and more severe for less educated women than for women with a high school degree or beyond. Diminishing occupational discrimination against black women appears to have been responsible for most of black women's improved occupational status over the period. However, the characteristics of black and white women converged substantially over the period, also contributing to the convergence in the occupations held by black and white women. ENDNOTES 1. For a similar description of trends in the occupations held by blacks and whites see R. Farley and W. Allen, The Color Line and the Quality if Life in America (New York: Russell Sage Foundation, 1987), pp. 256-82. 2. The job content of domestic service jobs and other service jobs may have been very similar. For blacks, however, household service may have had a connotation of their work under slavery and hence have had lower status. 3. The index of dissimilarity is defined as: (1/2) [Sigma] w(i) - b(i) i where b(i) is the proportion of black women who work in occupation i, and w(i) is the proportion of white women who work in occupation i. If black and white women were distributed equally across occupations, b(i)-w(i) would be zero for all occupations, and the value of the index would be zero. If black women and white women were in completely separate occupations, then for each occupation i, either b(i) or w(i), would be equal to zero, and thus the value of the index would be 1. 4. See table 5.4 5. U.S. Commission on Civil Rights, "technical Appendix," tables B.28-B.31. 6. Richard L. Rowan, The Negro in the Textile Industry (Philadelphia: Industrial Research Unit, Wharton School of Finance and Commerce, Univ. of Pennsylvania, 1970), pp. 61-64. 7. An example of a law preventing whites from teaching black pupils is one passed by the Atlanta city council in 1915 "stating that blacks were not to teach whites and whites were not to teach blacks." See S. Harley and R. Terborg-Penn, eds. The Afro-American Woman: Struggles and Images (Port Washington, N.Y.: National Univ. Publications, 1978) p. 47. 8. Heckman and Payner attribute black women's break-through into the textile industry to the implementation of Title VII of the 1964 Civil Rights Act and to Executive Order 11246. See James J. Heckman and Brook S. Payner, "Determining the Impact of Federal Antidiscrimination Policy on the Economic Status of Black: A Study of South Carolina," American Economic Review, vol. 79, no. 1 (1989), pp. 38-77. 9. App. B, tables B.10-B.12 show hypothetical occupational distributions for southern and nonsouthern black and white women assuming each group had the other group's characteristics. 10. See app. C. 11. See app. C. 12. For details on how these hypothetical occupational distributions were generated, see app. B. 13. If white women had black women's characteristics, the index of occupational dissimilarity would have been 56.0. Fewer than 12 percent of white women would have been domestic servants. 14. This chapter has only analyzed racial differences across relatively broad occupational categories. It is possible that the distributions across different types of jobs within broad occupational categories also varies by race. Furthermore, work places might also be segregated by race, either because there exist predominately black and predominately white firms or because there is physical segregation within firms. Thus, these measures of occupational segregation are underestimates of the true extent of segregation in the work place. 15. The value of the index of occupational dissimilarity comparing black women's 1940 occupational distribution with their 1980 occupational distribution is 68.3. See table 5.16. 16. The value of the index of occupational dissimilarity comparing the hypothetical distribution with the actual 1940 distribution is 26.5, whereas the value of the index of occupational dissimilarity comparing the actual 1940 and 1980 distributions is 68.3. Chapter 6 Accounting for Black-White Female Wage Differentials 1940-80: A Multivariate Analysis Chapters 4 and 5 identified several potential sources of the historical black-white wage gap and of the convergence in black and white women's wages in recent decades: racial differences in education, regional distribution, urban-rural mix, and occupations. In this chapter the statistical technique known as multiple regression is used to quantify the effects of these and other characteristics on the relative wages of black and white women over the 1940-80 period. Multiple regression permits the researcher to isolate the individual effects of various measured characteristics, such as education or region, on women's wages. Multiple regression results can be used to predict the wage that would be received by a woman with any given combination of measured characteristics. This chapter uses data from the 1940, 1960, and 1980 censuses. Separate regressions are estimated for black and white women for each year, because black and white women likely have different pay structures: for instance, in the presence of labor market discrimination against black women, black women probably receive lower wages than white women with the same characteristics. The regression results are then used to predict the wages black women would receive if they were paid according to the white pay structure and vice versa. Thus, the regression results allow comparisons between the wages earned by black women and the wages earned by white women with the same characteristics. They also allow comparisons between white women's wages and the wages black women would earn if black and white women had the same pay structures. The statistical analysis undertaken in this chapter, thus, provides the basis for answering the following questions central to uncovering the role played by racial differences in characteristics and hence the role played by labor market discrimination in determining the relative wages of black women: How much less did black women earn than white women with the same characteristics in each of the years 1940, 1960, and 1980? How much less than white women would black women have earned on the basis of their characteristics alone? Which characteristics of black women would have contributed to their lower earnings? Were there regional differences or differences across age groups or educational levels in black women's pay relative to comparable white women? What were the contributions of converging characteristics and converging pay structures to the increase in black women's relative pay between 1940 and 1980? Accounting for Black-White Wage Differentials: 1940, 1960, and 1980 This section looks at the extent to which wage differences between black and white women in 1940, 1960, and 1980 can be accounted for by racial differences in characteristics measurable with census data. The analysis is based on the results of estimating separate wage regressions for black and white women for each of the 3 years. Characteristics considered in the regression analysis include the following measurable factors thought to affect productivity: age, education, region, urban-rural residence, marital status, presence of children, full-time/part-time status, and (in some cases) occupation and industry of employment. (1) The Effect of Racial Differences in Characteristics and Pay Table 6.1, which reports results from wage regressions incorporating all of the characteristics enumerated above except for occupation and industry, allows consideration of the relative importance of racial differences in all measured characteristics taken together and of racial differences in pay structures as sources of the black-white wage gaps in 1940, 1960. and 1980. Panel A of table 6.1 shows the predicted hourly wages (in 1979 dollars) of black and white women with average characteristics for their group for 1940, 1960, and 1980. The first two lines of panel A show the wages the average black and white woman would have earned if they were paid according to the estimated black pay structure; the second two lines show the wages they would have earned if they were paid according to the estimated white pay structure. Thus, in 1940 the average black woman earned $0.87 an hour, and the average white woman earned $2.38 an hour. If the average black woman had been paid according to the white pay structure, she would have earned $1.73 an hour; if the average white woman had been paid according to the black pay structure, she would have earned $1.46 an hour. Panel B, line 1, reports the ratio of the average black woman's actual wage to the average white woman's actual wage. In 1940 the average black woman earned only 37 percent as much as the average white woman. This ratio increased to 55 percent in 1960 and reached 94 percent in 1980. Line 2 reports the ratio of the average black woman's wage to the average white woman's wage if both black and white women had been paid according to the white pay structure. Thus, line 2 shows the wage ratio that would prevail if black and white women had the same pay structure, or the ratio resulting from racial differences in measured characteristics alone. The lower the ratio in line 2, the larger the overall effect of racial differences in characteristics on the black-white wage ratio. Line 3 reports the ratio of the average black woman's wage to the wage she would have earned if she were paid according to the white pay structure (or the wage paid a white woman with the same measured characteristics). Thus, line 3 shows the wage ratio that would prevail if black and white women had the same characteristics, or the ratio resulting from differences in the pay structures facing black and white women alone. The lower the ratio in line 3, the larger the effect of racial differences in pay structures on the black-white wage ratio. In 1940 the average black woman earned only half as much as a comparable white woman. (3) Thus, differences in black and white women's pay structures were very large. Yet, even if black women had been paid according to white women's pay structure, the black- white wage ratio would have been substantially less than 1:72 percent. Thus, black women's characteristics would have prevented them from earning on a par with white women even if black and white women had been paid comparably. Differences in the pay received by comparable black and white women appear to have diminished slightly between 1940 and 1960. Black women in 1960 earned almost two-thirds as much as comparable white women. Nevertheless, differences in black and white women's pay structures continued to be an important source of the black-white wage gap. The characteristics of black and white women also became more similar between 1940 and 1960. In 1960 black women would have earned 86 percent as much as white women if there were no differences in black and white women's pay structures. Although racial differences in characteristics lowered black women's relative wages in 1940 and 1960, differences in black and white women's pay structures appear to have been the more important source of the black-white wage gap. The black-white wage ratios due to differences in pay structures alone (51 percent in 1940, 65 percent in 1960) were much lower than the black-white wage ratios due to differences in characteristics alone (72 percent in 1940, 86 percent in 1960). In 1980, however, differences in the pay structures for black and white women appear to have disappeared, leaving differences in characteristics as the sole source of the black-white wage gap. The average black woman in 1980 was paid 100 percent as much as a comparable white woman. Black-white differences in characteristics declined between 1960 and 1980, but black women's characteristics continued to lower their relative earnings in 1980. In 1980 differences in characteristics by themselves would have produced a black-white wage ratio of 94 percent. Since the actual black-white wage ratio in 1980 was 94 percent, differences in measured characteristics accounted for the entire black-white wage gap in 1980. In sum, there were large unexplained differences in the pay received by comparable black and white women in 1940 and 1960, suggesting that racial discrimination in the labor market may have been very important in limiting black women's earnings in those years. In 1980, however, there were virtually no differences in the pay received by comparable black and white women, and thus the census data do not by themselves provide evidence that labor market discrimination on the basis of race had a negative effect on black women's pay in 1980. Of course, if the average black woman had more work experience, a characteristic unmeasured by the census, than her white counterpart, she should have earned more than 100 percent as much. Thus, even though no difference between black women's pay and that of comparable white women can be detected using census data, it is possible that labor market discrimination continued to have adverse effects on black women's wages in 1980. (4) Racial differences in characteristics also contributed to black women's low relative earnings in all 3 years, but appear to have played a lesser role than differences in pay in 1940 and 1960. Both the characteristics of black and white women and the pay structures facing black and white women appear to have converged over the 1940-80 period. The Effect of Selected Characteristics on Black-White Wage Differentials Table 6.2 shows the separate effects of racial differences in selected characteristics on the black-white wage ratio. The first line of table 6.2 reports the actual hourly wage ratio for black and white women with average characteristics for their group. (5) The second line shows the ratio that would have existed if black women had been paid according to the white pay structure and represents the effect of racial differences in all characteristics taken together on the black-white wage ratio. The lower the ratio in line two, the larger the combined effect of characteristics as a group. Lines 3 through 8 show the black-white ratio that would have existed if black women had been paid according to the white pay structure and had average white values for all characteristics except for the selected characteristic but retained their own average value for the selected characteristic. This ratio measures the separate effect of racial differences in the selected characteristic on the black-white wage ratio. The lower this ratio, the larger the effect of the selected characteristic. Education As suggested by the analysis in chapter 4, black women's lower education had a substantial effect on the black-white wage ratio. Education was by far the most important characteristic limiting black women's wages in all 3 years. In 1940 black-white differences in education alone would have reduced black women's earnings by 19 percent. Consistent with the convergence in black-white schooling levels found in chapter 4, the effect of education diminished over time. In 1960 racial differences in education decreased black women's earnings by only 10 percent, and by 1980 by only 4 percent. The effect of education remained, however, larger than that of any other individual characteristic through 1980. Region Racial differences in regional distribution (the higher proportion of black women living in the South) had a small influence on the black- white wage ratio in 1940 and 1960 and virtually none in 1980: differences in regional distribution alone lowered black women's relative wage by 4 percent in 1940, by 3 percent in 1960, and by 1 percent in 1980. The small measured impact of region arises because the white pay structure was used to evaluate the effects of differences in characteristics on the black-white wage ratio. Had the black pay structure been used to evaluate the effect of differences in characteristics regional distribution would have appeared to have had a much larger influence on the black-white wage ratio. (7) This is because whites in the South were paid only slightly less than identical whites in the rest of the country (9 percent less in 1940 and 1960 and 5 percent less in 1980), whereas blacks in the South were paid much less than identical blacks outside of the South (45 percent less in 1940, 44 percent less in 1960, and 18 percent less in 1980). (8) The large wage penalty associated with living in the South for blacks may well be due primarily to greater labor market discrimination against blacks in the South than in the rest of the country. Since the goal here is to evaluate the effects of racial differences in characteristics separately from the effects of discrimination, it is preferable to use the white pay structure to evaluate the influence of racial differences in regional distribution alone on the black-white wage gap. Urban or Rural Location Racial differences in proportions living in urban areas accounted for a small portion of the wage gap in 1940, lowering black women's relative wage by roughly 3 percent, but had no effect on the black-white wage ratio in the later years. This is consistent with the findings in chapter 4 that a higher proportion of black women lived in rural areas in 1940, but that by 1960 the reverse was true. Moreover, urban wages were much higher than rural wages at the beginning of the period, but that wage advantage declined steadily over the period. Urban women earned 29 percent more than their rural counter-parts in 1940, 19 percent more in 1960, and 13 percent more in 1980. (9) Other Characteristics Racial differences in demographic factors (marital status and number of children) had a small effect on the black-white wage ratio in 1940, but none in 1960 and 1980. The higher proportion of black women working part time actually raised the black-white wage ratio in 1940, but had no effect in the later years. The Role of Occupation and Industry Occupation and industry were omitted in the regressions on which table 6.1 and 6.2 are based, because, unlike the other characteristics considered, the occupations and industries in which women are employed can be directly affected by labor market discrimination. Indeed, chapter 5 argues that at least in 1940 and 1960, labor market discrimination probably substantially limited occupational opportunities for black women. As a result, including occupation and industry among the characteristics incorporated in a wage regression could lead to a false conclusion that black-white wage differences are "explained" by differences in characteristics, and hence to underestimation of the extent of labor market discrimination. It is useful, however, also to consider wage regressions incorporating occupation and industry, because these regressions can yield important information about the nature of the differences in the pay structures of black and white women. Using results from these regressions to determine the influence of racial differences ln occupations and industries on the black-white wage ratio makes lt possible to answer the following question: Did black women earn less than white women within occupations and industries. or did they earn less than white women because they were in low-paying occupations and industries? The discussion here is based on tables 6.3 and 6.4. Table 6.3 is similar to table 6.1 but is based on regressions that include occupation and industry in addition to the characteristics included in the regressions used in table 6.1. Table 6.4 corresponds to table 6.2. Comparisons of tables 6.3 and 6.4 with tables 6.1 and 6.2 suggest that a major reason why black women were paid less than white women was that they worked in occupations and industries that paid less than those that employed comparable white women. When occupation and industry are included among the characteristics in the regression analysis, a much larger portion of the black-white wage gap appears to be "explained" by racial differences in characteristics. In 1940, for instance, when occupation and industry are omitted from the regression analysis, characteristics can explain very little of the black-white differences in pay. Compared to the actual 37 percent black-white wage ratio, racial differences in all characteristics taken together would have produced the much higher wage ratio of 72 percent if there were no racial differences in pay structures. When occupation and industry are included among the regression characteristics, however, characteristics explain most of the wage gap, producing a ratio of 47 percent. Moreover, racial differences in occupation and industry alone would have produced a black-white wage ratio of 57 percent in 1940 (see table 6.4), indicating that occupation and industry were by far the most important "characteristic" limiting black women's earnings in that year. Racial differences in occupation and industry continued to be important factors limiting black women's earnings in 1960, when they would have been responsible for a black-white wage ratio of 70 percent, but their importance had declined substantially by 1980, when they would have produced a wage ratio of 96 percent. Incorporating occupation and industry in the regression analysis reduces (but does not entirely eliminate) the apparent effect of edu- cation on the black-white wage ratio: ln 1940, for instance, the wage ratio produced by black-white differences in education alone increases from 81 percent to 91 percent when occupation and industry are added to the regression characteristics. Similar results are obtained in the other years. Thus, it appears that the negative effect of racial differences in education on black women's relative wages observed in table 6.2 occurs partly because black women's lower education causes them to enter lower paying occupations and industries and partly because their lower education causes them to earn less than white women within occupations and industries. Regressions incorporating occupation and industry leave much less of the black-white pay gap "unexplained" than regressions that omit them, and thus they appear to imply a much smaller potential role for labor market discrimination. However, since the occupations and industries women work in can themselves be affected by labor market discrimination--indeed, chapter 5 suggests that labor market discrimination may have severely limited black women's occupational opportunities--it is probable that labor market discrimination played a much more important role than suggested by table 6.3. It seems likely that the estimates of discrimination derived from table 6.1 better reflect the true extent of discrimination. Pay Differences by Age, Education, and Region Earlier, this chapter noted that black women earned considerably less than comparable white women in 1940 and 1960 and about the same amount in 1980. Here the analysis is refined to consider whether black women's pay differs relative to comparable white women's across age groups, educational levels, and regions. Age To determine whether black women's pay relative to comparable white women differed for younger and older women, black and white wage equations were estimated separately for women under 40 years old and over 40 years old in 1940, 1960, and 1980. Table 6.5 presents predicted wages and black-white wage ratios for younger and older women derived from regressions that do not include occupation and industry among the regression characteristics. (10) In 1940 and in 1960 racial differences in characteristics lowered the relative wages of younger black women less than they did those of older black women. This result suggests that at least with respect to measured characteristics younger black women had better relative skill levels than older black women and is consistent with findings in chapter 4 that black-white education differentials were lower among younger women. By 1980, however, there were no apparent age differences in the relative skill levels of black women. In 1960 and 1980 younger black women earned more relative to white women in their age group than did older black women. One reason why younger women earned relatively more in 1960 and 1980 is that they were better paid relative to comparable white women (panel B, line 3). When occupation and industry are included in the regressions (app. D, table D.5), on the other hand, the relative advantage of younger black women disappears, indicating that within occupations they were paid the same as their older counterparts. Thus, the main reason why younger black women earned more relative to comparable white women is that they had relatively better occupational opportunities than their older counterparts. These results suggest that older black women may have fared worse because they continued to feel the effects of greater past discrimination: they may have been in relatively lower paying occupations because they had made their initial occupational choices at a time when labor market discrimination was more severe and black women's access to occupations more limited. Additionally, older black women's occupational opportunities may have been limited because they differed more from their white counterparts in unmeasured characteristics, such as the quality (as opposed to quantity) of education. This explanation is consistent with the trend of increasing quality of black schooling noted in chapter 4. Education Table 6.6 presents results by educational attainment instead of by age. The results in table 6.6 are based on black and white wage equations estimated separately for women with 11 years of education or less, for women with exactly 12 years of education, and for women with 13 or more years of education in each of the years 1940, 1960, and 1980. (11) In all 3 years, black women earned more relative to white women at higher educational levels. In 1940 this effect showed up only for women with some college education, but in 1960 and 1980, black women's relative wages rose continuously with education. The main reason why more educated black women earned relatively more is that there were smaller racial differences in pay at higher educational levels. These results suggest that labor market discrimination may have had a larger negative effect on the wages of less educated black women. One possible explanation is that highly educated black women had relatively wider occupational opportunities, such as the possibility of entering the teaching profession (see chapter 5). Indeed, when occupation and industry are included in the wage regressions (appendix D, table D.6), the relative advantage of educated black women disappears. This implies that within occupations and industries, highly educated black women were just as poorly paid relative to white women as less educated black women were, but that they had better occupational opportunities relative to equally educated white women than did their less educated counterparts. A second possible explanation for the relative success of highly educated black women is the following. Because comparatively few black women obtain a college education, those who do may be particularly able individuals and earn more for this reason. A third possibility is that highly educated white women themselves faced a greater degree of sex discrimination in the labor market than white women with less education. The more similar wages and occupations of educated black and white women could, then, be an indication of greater discrimination against white women at higher educational levels rather than of lesser discrimination against black women at higher educational levels. The racial differences in pay structures diminished over time at all three educational levels. Between 1940 and 1960, educated black women made larger gains, and after 1960 less educated black women made larger gains. Thus, the effects of discrimination may have abated first for highly educated black women and later for less educated black women. This is consistent with the finding in chapter 5 that occupational opportunities opened up for educated black women first, and for less educated black women later Region Table 6.7 presents the results by region (South/non-South). (12) In all 3 years. black women earned considerably less relative to white women in the South than in the rest of the country. In 1940 the average black woman earned.just over one-third as much as the average white woman in the South, but in the rest of the country she earned more than two-thirds as much. Regional differences in pay declined over time but did not entirely disappear: in 1980 the average southern black woman earned only 87 percent as much as southern white women, whereas outside of the south black women earned 105 percent as much as white women. One reason for southern black women's lower relative pay was that, particularly in 1940, they were relatively less skilled than black women in the rest of the country. Even if black women had been paid according to the white pay structure in 1940, southern black women would have earned only 71 percent as much as southern white women, whereas black women outside of the South would have earned 91 percent as much as white women outside of the South (panel B, line 2). The regional skill differential narrowed over time. By 1960 southern black women's relative wages would have increased from 71 to 87 percent if they had been paid according to the white pay structure, compared with a much smaller increase from 91 to 94 percent outside of the South. However, even in 1980, southern black women were relatively less skilled than black women in the rest of the country. Even more important than their relatively low skills, unexplained racial differences in pay structures were greater in the South than in the rest of the country. In 1940 southern black women earned only 53 percent as much as comparable white women, whereas black women outside of the South earned 71 percent as much as their white counterparts (panel B, line 3). Between 1940 and 1960, black women's wage did not increase relative to comparable white women in the South, but they did increase in the rest of the country. By 1980 black women's wage relative to comparable white women had increased in the South as well, but southern black women continued to earn less than white women with the same characteristics, whereas, outside of the South, black women had achieved wage parity with comparable white women. The relative disadvantage of black women in the South does not disappear when the analysis is repeated using regressions incorporating occupation and industry (see appendix D, table D.7). Thus, not only did black women in the South have more restricted occupational opportunities than black women in the rest of the country (see chapter .5), but they were also paid less relative to identical white women within occupations and industries than were black women in the non-South. These results suggest that labor market discrimination may have had and may continue to have a greater negative effect on black women living in the South than on black women living in the rest of the country, although regional differences in black women's unmeasured characteristics, such as quality of education, could be partially responsible. The effects of labor market discrimination against black women also appear not to have begun to diminish in the South until after 1960, whereas they began to diminish earlier in the rest of the country. The analysis in this section suggests that the effect of discrimination was not felt evenly by black women. Older black women have generally fared worse than younger black women, probably because they cannot entirely overcome the legacy of past discrimination. Highly educated black women appear to have experienced less wage discrimination on the basis of race than black women at lower educational levels. This may have been because of the relatively wider occupational opportunities open to the few black women who were highly educated, or alternatively because of greater sex discrimination against educated white women. Black women in the South appear to have suffered greater wage discrimination than black women in the rest of the country. Accounting for the Convergence in the Black-White Female Wage Ratio: 1940 to 1980. Between 1940 and 1980, the black-white female hourly wage ratio increased from 37 percent to 94 percent. This section investigates the extent to which convergence in the characteristics of black and white women can account for this remarkable increase. To do so, the 1940 and 1980 regressions were used to predict the wages black and white women would have earned in 1940 if they had had their 1980 characteristics, and the wages they would have earned in 1980 if they had had their 1940 characteristics. The results for regressions that do not include occupation and industry are presented in table 6.8. Line 3 of table 6.8 measures the impact of changing pay structures on women's wages and on the black-white wage ratio. Even if the characteristics of black and white women had not changed between 1940 and 1980, the black-white female wage ratio would have more than doubled, increasing from 37 percent to 77 percent. Line 4 measures the impact of changing characteristics on women's wages and on the black-white wage ratio. If pay structures had not changed between 1940 and 1980, the black-white wage ratio would have increased by 16 percentage points, from 37 percent to 53 percent, due to changing characteristics. Thus, although the converging characteristics of black and white women contributed to the increased black-white wage ratio between 1940 and 1980, converging pay structures were much more important as a source of black women's increasing relative wages. Lines 5 through 11 measure the effects of changes in individual characteristics on wages and on the black-white wage ratio. For example, a black woman in 1940 who had the 1940 black average for all characteristics except for education and the 1980 black average education would have earned $1.25 an hour instead of the $0.87 an hour she would have earned if she had had the 1940 black average education. If both black and white women in 1940 had their group's 1940 average for all characteristics except for education and their group's 1980 average education, the black-white wage ratio would have been 44 percent instead of the actual value of 37 percent. Of the characteristics shown, the only ones that can account for more than 1 percentage point of the increase in the black-white wage ratio between 1940 and 1980 were education, regional distribution. and urban/rural residence. Of these, education accounts for the largest increase. Table 6.9 is similar to table 6.8, but it is derived from the regressions that include occupation and industry. When occupation and industry are included among the regression characteristics, convergence in the characteristics of black and white women accounts for more of the increase in the black-white wage ratio than when they are not, and indeed, it accounts for more of the increase than does convergence in the pay structures for black and white women. Occupation and industry by themselves account for an in- crease in the black-white wage ratio from 37 percent to 54 percent, or more than one-half of the increase. Education, regional distribution, and urban-rural mix also continue to have small effects. Overall, the racial convergence in characteristics other than occupation and industry over the 1940-80 period accounts for less of the increase in the black-white wage ratio than does convergence in the pay structures for black and white women. The pay structures for black and white women appear to have converged partially because the occupational and industrial distributions of black and white women became more similar and partly because black women's relative wages increased within occupations and industries. Since declining discrimination over the 1940- 80 period was likely responsible for much of both the convergence in the black-white pay structures and the convergence in the occupational and industrial distributions of black and white women, these results appear to indicate that it was declining labor market discrimination and not changes in the characteristics of black women that caused most of the increase in their relative wage between 1940 and 1980. Summary and Conclusions Although characteristics not including occupation and industry can account for some of the black-white wage gap in 1940 and 1960, differences in the pay structures for black and white women were a far more important cause of black women's low relative wage in those years. In 1940 the average black woman earned just half as much as she would have if she were white, and In 1960 she earned two-thirds as much. In 1980, on the other hand, differences in pay structures appear to have been unimportant: the average black woman earned the same wage she would have if she were white. The small black-white wage gap that remained in 1980 can be fully accounted for by racial differences in characteristics. In all 3 years, education was the most important among the characteristics limiting black women's earnings. Black-white differences in occupations and industries accounted for much of their differences in pay structures in 1940 and 1960. Black women earned less than white women with the same characteristics both because they entered low-paying occupations and industries and because they were paid less than identical white women within occupations and industries. The black-white differences in occupations and industries that remained in 1980 had only a very small effect on the black-white wage ratio. The results suggest that racially motivated labor market discrimination against black women probably lowered their earnings substantially in 1940 and 1960, both by lowering their occupational attainment and by lowering their pay within occupations and industries. The effects of current racial discrimination most likely had diminished considerably by 1980. However, black women over 40 may have continued to suffer the effects of past discrimination: their current occupational status and hence their wages may have continued to be lowered because they had experienced limited occupational opportunities at the outset of their labor market careers. Labor market discrimination appears to have affected black women unevenly. Black women with low educational levels and black women living in the South earned less relative to comparable white women than did more educated black women and black women living outside of the South. Both racial convergence in characteristics and convergence of the pay structures of black and white women contributed to the significant increase in the black-white female wage ratio from 37 percent in 1940 to 94 percent in 1980. Convergence of the pay structures appears to have been the more important of the two. Thus, declining labor market discrimination against black women was probably the most important reason for their increased relative wage over the period, although changes in characteristics, such as educational attainment and region of residence, also played a role. Labor market discrimination appears to have declined along two fronts: the relative occupational opportunities for black women improved significantly over the 1940-80 period and the relative pay received by black women within occupations and industries also increased. It should be remembered, however, that the conclusions in this chapter are based on regression analysis of census data and thus are only as good as that data. For several reasons, census data may permit only an incomplete analysis of black-white wage differentials. First, the census wage and salary data reported do not include any in-kind payments a worker may have received. Yet, large fractions of black women (and almost no white women) in 1940 and in 1960 were employed either as domestic servants or as farm laborers (many on family farms), both of which are occupations in which in-kind payment has traditionally represented a sizable portion of total compensation. If in-kind payments were Incorporated in the wage and salary data used in this analysis, the estimated relative wages of black women in 1940 and 1960 would undoubtedly have been higher and the estimated degree of discrimination lower than those derived in this chapter. Consequently, the degree of wage improvement experienced by black women over the years 1940-80 may have been less than census wage data indicate, since this was a period during which black women left agricultural and domestic service jobs for jobs less likely to provide in-kind pay. (13) Second, just as census measures of wages might be inaccurate, so might census measures of the characteristics included in the regression analysis. In particular, as we indicated in chapter 5, the census data on years of schooling completed might not reflect accurately women's actual educational achievement. The census data probably overestimate the relative educational attainment of black women, particularly in 1940 and 1960, making it appear closer to that of white women than it actually was. If so, the analysis in this chapter might overstate the degree of labor market discrimination against black women in those years. Third, the censuses do not collect data on the previous work experience of workers. Many researchers have found that workers with more work experience have acquired better labor market skills and earn higher wages as a result. Moreover, it is likely that during the 1940-80 period black women worked more continuously than otherwise similar white women and hence accumulated more work experience (and more labor market skills). If this is the case, then in the absence of discrimination, a black woman would earn more than a white woman with the same characteristics (not including experience). Excluding work from the characteristics incorporated in the regressions in this chapter could, thus, lead to underestimation of the degree of discrimination against black women. The regression analysis used in this chapter only yields estimates of the direct effect of labor market discrimination on black-white wage differentials. It is likely that labor market discrimination also has indirect effects on black-white wage differentials. For in- stance, black women might be discouraged by the prospect of discrimination and as a result acquire fewer skills or less education than white women. The analysis used in this chapter would attribute this type of indirect effect to racial differences in characteristics and not to discrimination. Moreover, discrimination not directly related to the labor market may well affect the characteristics of black women and hence lower their relative wages. For instance, discrimination in the provision of schooling may limit the education received by black women. Again, the methodology used in this chapter attributes the effects of this type of discrimination to racial differences in characteristics. As a result, it is possible that the overall effect of discrimination on black women's relative wages is much larger than the direct effect of labor market discrimination estimated in this chapter. Finally, it should be remembered that white women, too, are subject to labor market discrimination. The methodology employed in this chapter only allows measurement of the added effects of discrimination faced by black women. Despite its shortcomings, the census is the only source of consistent data on the wages and characteristics for a representative sample of the population going back to 1940. Although the estimates of discrimination derived from the census data may not be precise, the census data do provide a broad picture of trends between 1940 and 1980 that is unavailable from any other data source. For the contemporary period, fortunately, data sources exist that overcome many of the deficiencies of the census data. The next chapter relies on the Survey of Income and Program Participation (SIPP) to generate estimates of labor market discrimination against black women for the contemporary period. The main advantage of the SIPP over census data for this purpose is that the SIPP provides information on women's work experience and, thus, more accurate measures of women's labor market skills. In addition, the SIPP data are collected by trained interviewers, which most likely improves the accuracy of the data on other characteristics. ENDNOTES 1. The regressions used in this chapter are shown in app. D. 2. "Actual" wage for each group here refers to the predicted wage for the average woman from the group using the estimated pay scale for that group. 3. The term "comparable white women: refers here and throughout the report to a white woman with the same measured characteristics as the average black woman. 4. The next chapter uses a different data source, the Survey on Income and Program Participation, to look at the effect of racial differences in labor market experience on measures of labor market discrimination againast black women. 5. This line is the same as line 1, panel B, in table 7.1. 6. This line is the same as line 2, panel B, in table 7.1. 7. See app. table D.3, which reproduces table 7.2 using the black pay structure instead of the white pay structure to evaluate the effect of differences in characteristics. 8. The wage differences between southern and non-southern women cited here hold constant other differences between southern and nonsouthern women. They are the coefficients on a dummy variable for living in the south in the census wage regressions for black and white women reported in app. D, table D.1. 9. These figures hold constant other differences between urban and rural women. They are the coefficients on an urban dummy variable in the census wage regressions for white women. Se app. D, table D.1. 10. Predicted wages and black-white wage ratios by age for regressions that do include occupation and industry are reported in app. D, table D.5. 11. App. D, table D.6, presents comparable results for regressions including occupation and industry. 12. Results for regressions including occupation and industry are shown in app. D, table D.7. 13. Claudia Goldin (unpublished) has suggested that taking in0kind payments into account might increase the 1940 black-white female annual earnings ratio for full-time year-round workers from 37 percent to 49 percent. If so, the situation of black women in 1940 may have been substantially better than suggested by the unadjusted census figures. On the other hand, the pace of the post-1940 growth of black women's earnings would then be slower than implied by the unadjusted census figures. Chapter 7 Accounting for the Black-White Female Wage Differential in the 1980s: A Multivariate Analysis This chapter updates to the 1980s and extends the 1940-80 wage analysis of chapter 6 by using the Survey of Income and Program Participation (SIPP). To make intertemporal comparisons, the SIPP data are supplemented with data from the March 1980, March 1985, and March 1987 Current Population Surveys. The SIPP contains information on a large nationally representative sample of men, women, and children. The SIPP data have several advantages over the census data employed in chapters 4 through 6. The SIPP is an extremely rich source of data on all aspects of persons' economic and demographic status. It provides detailed information on persons' wages, income, labor force status, and family structure on a monthly basis for 32 months. It also provides rich information on a number of topics, including family assets, and retrospective information on persons' education and work force, marital, and fertility histories. All members of a household are included in the sample, allowing data on women to be merged with data on their husbands, their children, and other persons living with them. Particularly useful for analyzing racial differences in wages is the SIPP's information on the previous work experience of women. In addition, the SIPP wage data cover a more recent period than currently available through the census data: July through November 1984. Because they are based on data from a different source, however, the results obtained from SIPP data cannot be compared directly with results for earlier years and thus are of limited value in making intertemporal comparisons. To supplement the SIPP data, this chapter also uses data from the March 1980, March 1985, and March 1987 Current Population Surveys (CPS). The March 1980 CPS covers approximately the same period as the 1980 census; the March 1985 CPS covers the same period as the SIPP; and the March 1987 CPS covers an even more recent period (1986). Because the CPS data are directly comparable across years, they allow intertemporal comparisons to assess the degree of change during the 1980s. Moreover, the March 1980 CPS data can be compared with data from the 1980 census, and the March 1985 CPS data can be compared with data from the SIPP, allowing comparison of results from the three data sources. The CPS data share the main drawback of the census data, however, which is that they do not provide information on women's previous work experience. Accounting for Black-White Wage Differentials in the 1980 This section reports the general results of the statistical analysis of women's wages in the 1980s using the SIPP. As in Chapter 6, the approach is to measure the separate effects of racial differences in characteristics and of racial differences in pay structures on the black-white wage ratio. Black-white wage differentials due to racial differences in characteristics alone can be assumed not to be the direct result of labor market discrimination against black women.(1) On the other hand, black-white wage differentials due to racial differences in pay structures may reflect labor market discrimination against black women. (The reader is referred to chapter 6 for a detailed discussion of the statistical methodology employed in this chapter.) This section highlights the importance of including variables pertaining to women's previous work experience among the characteristics considered in the regression analysis of wages. Many researchers have found that persons with more overall work experience command higher wages, probably because they have accumulated more job-related skills. Similarly, persons with greater seniority on their current job (job tenure) have also been found to command higher wages, possibly due to their accumulation of skills relevant to their specific job. Some researchers have found that intermittent work causes labor market skills to atrophy, and hence persons who have spent more years out of the work force receive lower pay than others with similar characteristics.(2) Table 7.1, showing means of work experience variables by race for the SIPP data, confirms that black women have more overall work experience and greater seniority in their current jobs as well as less time out of the work force (home time) than white women. As a result, it is important to include variables pertaining to work experience, job seniority, and time out of the work force in a statistical analysis of black-white female wage differentials.(3) The analysis here is based on wage regressions estimated separately for black and white women using the SIPP data. Wage regressions were estimated both with and without variables pertaining to women's work experience to assess the effect of racial differences in work experience on black-white women's wage differentials. A third specification, which statistically corrects for "selectivity bias," or the possibility that the wages of women who are actually working are biased indicators of the wage that would be received by women as a group, including women who are not currently working, was also implemented.(4) Besides variables pertaining to women's work experience, the characteristics included in the SIPP wage regressions are: education, full-time/part-time status, urban or rural residence, region, marital status, and presence and number of young children, and in some cases, occupation and industry. The SIPP regression results are presented in appendix F, tables F.1 and F.2. Following the methodology developed in chapter 6, predicted wages for black and white women and predicted black-white wage ratios are presented for the SIPP regressions. Table 7.2 shows results for regressions that do not include occupation and industry, and table 7.3 shows results for regressions that do include occupation and industry. For both tables column (1) shows results when work experience variables are not included and selectivity bias is not corrected for, column (2) shows the results when work experience variables are added, and column (3) shows results when, in addition, selectivity bias is corrected for. According to the SIPP data, the average black woman earns approximately 90 percent as much as the average white woman. When women's occupations and industries are not taken into account (table 7.2), the SIPP regressions that do not include variables pertaining to women's work experience (column (1)) suggest that differences in the characteristics of black and white women are partially responsible for black women's lower wages--the average black woman is predicted to earn approximately 3 percent less than the average white woman even if both groups were paid according to the white pay structure. However, contrary to results from the 1980 census, differences in the pay structures facing black and white women appear to be an equally important source of their lower earnings: the average black woman earns approximately 5 percent less than a comparable white woman. When work experience variables are entered into regressions using the SIPP data, whether or not the regressions control for selectivity bias (columns (2) and (3)), racial differences in characteristics account for even less of the wage differential--the average black woman is predicted to earn only 1 percent less than the average white woman if both groups were paid comparably. Differences in the pay structures for black and white women appear to account for most of the wage differential-- the average black woman earns only 9 percent less than a white woman with the same characteristics. When women's occupations and industries are taken into account (table 7.3), a larger portion of the black-white wage differential is accounted for by characteristics and less is unexplained. The SIPP regression that excludes work experience variables predicts that the average black woman earns almost 100 percent as much as a comparable white woman. When work experience variables are included, this figure drops to 96 percent. These results imply that a major reason why black women continue to earn less than white women is that they work in lower paying occupations and industries. Why black and white women continue to work in different jobs is an important issue that remains to be explored. The explanation may lie in part with the lingering effects for older black women of occupational discrimination in the decades before to the 1964 Civil Rights Act. Another possible explanation is that racial differences in unmeasured productivity-related characteristics restrict black women's relative job opportunities. Finally, it is possible that many black women continue to face occupational discrimination in today's labor market. The results in tables 7.2 and 7.3 point to the importance of including variables pertaining to work experience when analyzing black-white female wage differentials. In each case, including these variables lowers by around 4 percentage points the estimates of what the average black woman earns relative to a comparable white woman. This suggests that analyses of the black-white female wage gap that do not take into account racial differences in work experience may underestimate the degree of labor market discrimination against black women. In particular, the results in chapter 6 should be reevaluated in this light. Fortunately, these results suggest that, at least for the current period, the bias produced by not including data on work experience is relatively small. The results in tables 7.2 and 7.3 suggest that the characteristics of black women (other than occupation and industry) no longer lower their relative earnings significantly: estimates based on the SIPP data suggest that if there were no racial differences in pay structures, black women would earn almost 99 percent as much as white women.(5) Despite the significant progress made by black women between 1940 and 1980, however, the results suggest that lingering racial discrimination in the labor market may continue to lower the relative earnings of black women: the average black woman is estimated to earn only 91 percent as much as she would if she were white. When occupations and industries are included in the analysis, this figure rises to 96 percent, indicating that slightly over half of the wage differential between identical black and white women arises because the black women work in lower paying occupations and industries. It is necessary, however, to temper these conclusions slightly. Although the wage regressions using the SIPP data improve on the regressions using the census data by including variables pertaining to work experience, the SIPP wage regressions themselves may omit other productivity-related characteristics that vary by race. To the extent that productivity-related characteristics are omitted, this chapter may either underestimate or overestimate the direct effect of discrimination. Furthermore, as discussed above, the methodology used in this chapter does not quantify any indirect effects of labor market discrimination. Pay Differences by Age, Region, and Education This section investigates whether black women fare differently by age group, across regions of the country, and by educational level. Separate regressions are estimated for each age group, for each region, and for each education level. All of these regressions use SIPP data, include variables pertaining to work experience, and correct for selectivity bias. Regressions were estimated both with and without occupation and industry. Age Whereas older black women were educated in impoverished, segregated black school systems, younger black women are likely to have benefitted from attending better schools. In addition, older black women are likely to have faced greater discrimination when they started their labor market careers. By constraining their early career choices, this early discrimination may have continued to lower the relative earnings of older black women later in life. To address this issue, separate regressions are estimated for women under and over 40 years of age.(6) Table 7.4 presents the results. Both older and younger black women earn approximately 90 percent as much per hour as their white counterparts when characteristics are not taken into account. After adjusting for characteristics, however, older black women are found to be at a greater disadvantage than younger black women: black women over 40 earn 88 percent as much as comparable white women, 6 percentage points below the corresponding figure for younger black women (94 percent). When occupation and industry are included among the regression characteristics, however, older black women's relative disadvantage disappears. Older black women actually earn more relative to identical white women than younger black women do: 99 percent versus 95 percent. Taken together, these results imply that, compared to similar white women, older black women are in lower paying occupations and industries than are younger black women, but within occupations and industries, older and younger black women earn comparably or even more. One possible explanation for these findings is that older black women have not entirely overcome the discriminatory constraints they faced in their early careers: having started working in lower paying occupations and industries because of limited opportunities early on, they have not entirely succeeded in leaving them. Yet within these occupations and industries, they earn almost as much as comparable white women. Thus, in addition to current discrimination, older black women may suffer from the lingering effects of past occupational discrimination. Region Previous chapters showed persistently larger black-white wage and occupational differentials in the South than in the rest of the country. Table 7.5, which presents results by region (South and non-South), indicates that southern black women do indeed continue to face a relative disadvantage. In the South the average black woman earns 83 percent as much as the average white woman, compared to 101 percent for the rest of the country. This result persists even after racial differences in characteristics are taken into account: the average black woman earns 87 percent as much as a comparable white woman in the South compared with 96 percent for the rest of the country. When women's occupations and industries are taken into account, however, black women in the South appear to earn as much as comparable white women. This result implies that the reason southern black women are paid less than southern white women is not because of racial differences in measured characteristics other than occupation and industry and not because black women are paid less than white women within occupations and industries but because they work in lower paying occupations and industries than their southern white counterparts.(7) Indeed, as indicated in chapter 5, the relative occupational status of black women remains lower in the South than in the rest of the country. Southern black women are under-represented among professional and clerical workers, and overrepresented among operatives, service workers, and domestic servants to a much greater degree than black women outside of the South. The extent to which Southern black women are under-represented among clerical workers is particularly striking: according to the SIPP data, 18 percent of southern black women are clerical workers, compared with 32 percent of southern white women and 32 percent of black women and 34 percent of white women in the rest of the country. As discussed in chapter 5, the relatively low occupational status of southern black women cannot be explained by measured characteristics. A possible explanation, which needs to be investigated, is that there continue to exist discriminatory obstacles for southern black women in some occupations, most notably in the clerical sector. An alternative possibility is that racial differences in unmeasured productivity-related characteristics limit black women's relative occupational status in the South. Whereas the relatively low occupational status of southern black women appears to be the source of black women's low relative wages in the South, racial differences in occupations and industries appear to play a much lesser role in the rest of the country. Nevertheless, even though black women earn on a par with white women outside of the South, their earnings may be lowered by discrimination: The average black woman would earn somewhat more than, not the same amount as, the average white woman if she were paid according to the white pay structure. In sum, black women continue to fare worse in the South than in the rest of the country. This research suggests that this is primarily because they work in lower paying occupations and industries than their white counterparts. Further research needs to be undertaken to determine whether black women's lower occupational status in the South is the result of labor market discrimination or whether it is due to some other cause. While black women outside of the South fare relatively better, they, too, earn less than predicted on the basis of their characteristics and thus might suffer from racial discrimination. Education Chapter 6 found that black women fared better relative to white women at higher education levels during the 1940-80 period. Table 7.6 shows separate regressions for women who are not high school graduates, for women with exactly 12 years of education and for women with at least some postsecondary schooling. Following the historical trend, black women continue to earn more relative to white women at higher educational levels: the average black woman with some college education earns 97 percent as much as her white counterpart, compared with 90 percent for women in the two lower educational groups. Contrary to results obtained using the 1980 census data, however, black women earn less than white women at all educational levels. Table 7.6 suggests that based on their characteristics, black women with 12 years or more of school should earn more than white women in their educational category, and black women with less than 12 years of school should earn almost as much when occupation and industry are not included among the regression characteristics. Yet, at all schooling levels black women earn less than comparable white women. Earning only 88 percent as much as comparable white women, black women with exactly 12 years of schooling appear to fare particularly poorly. When occupation and industry are included among the regression characteristics, black women's predicted relative wages based on their characteristics are lower. They also appear to earn more relative to comparable white women at all schooling levels, indicating that, at all schooling levels, black women work in lower paying occupations and industries than comparable white women. This is particularly true for black women with less than 12 years of school: when occupation and industry are included among the regression characteristics, black women in the lowest educational group actually earn more than white women with the same characteristics. Thus, two findings emerge from table 7.6. First, black women with exactly 12 years of school fare worse relative to similar white women than black women in the other educational groups. Whether this is due to differences in unmeasured characteristics, such as quality of education, or greater discrimination against these women is a matter for further research. Second, while the lower pay of black women in the two higher education groups is primarily due to lower pay within occupations and industries, this is not the case for black women with less than 12 years of education. Black women at the lowest educational level earn less than similar white women because they work in lower paying occupations and industries, not because they are paid less within occupations and industries.(8) The results in this section show that black women earn less than white women, both before and after taking racial differences in characteristics into account, both in the South and in the rest of the country, at all educational levels, and in both age groups.(9) Southern black women, black women over 40 years of age, and black women with exactly 12 years of education appear to fare particularly poorly relative to their white counterparts. The lower relative wage of women with exactly 12 years of education results from differences in pay within occupations and industries. The lower relative wages of southern black women, of black women with less than 12 years of education, and of older black women appear to be due to racial differences in occupations and industries in which women work. For older black women, this appears to be the result of past rather than of current labor market discrimination. The reasons why black women in the South and black women with less than a high school education work in lower paying occupations and industries than comparable white women are not as readily apparent. Current labor market discrimination may continue to limit the job opportunities of southern black women and of black women with low educational attainment. Trends in Black-White Female Wage Differentials over the 1980s Although the SIPP provides better information about women's characteristics at a single point in time (1984) than other data sources, it does not allow consideration of trends in women's wage differentials over time during the 1980s. For this purpose March Current Population Survey (CPS) data for the years 1980, 1985. and 1987 (covering the years 1979, 1984, and 1986, respectively) were used. Tables 7.7 and 7.8 report predicted wages and wage ratios for black and white women derived from regressions using the 1980, 1985, and 1987 CPS. Since the CPS does not provide information on women's work experience, women's ages were included in the regressions as a proxy for their work experience.(10) Other characteristics included ln the regression analysis are: education, full-time/part-time status, urban-rural residence, region, and marital status. For each year, regressions incorporating occupation and industry in addition to the characteristics enumerated above were also estimated.ll The CPS data suggest that the black-white wage ratio was constant at around 96 percent during the early part of the 1980s but fell to around 93 percent between 1984 and 1986. This was not due merely to a relative deterioration in the labor market skills of black women, although the predicted black-white wage ratio due to characteristics alone fell slightly between 1984 and 1986 (from 98 to 97 percent). Instead, there appears to have been a larger decrease in the ratio of the wage earned by the average black woman to that earned by a comparable white woman, whether or not occupation and industry are included among the regression characteristics. Thus, it appears that less of the black-white wage differential could be accounted for by racial differences in characteristics in 1986 than in 1984. It remains possible, however, that the 1984-86 decrease in black women's relative wages can be accounted for by changes in their levels of work experience, job tenure, and time out of the work force compared to those of white women. Unfortunately, until later panels of the SIPP are released, it is not possible to investigate this hypothesis. This apparent decrease in black women's relative wages occurred at around the same time that the female-male wage ratio for whites began to rise noticeably for the first time since World War II. Some have argued that the increase in the white female-male wage ratio is the result of declining gender discrimination in the labor market. If so, it would appear that black women have not benefitted to the same extent as white women from the decline in gender discrimination. An alternative explanation for the increase in the female-male ratio is that the labor market skills of white women improved during this period. For instance, following the long term increase in labor force participation by white women, it is likely that the work experience of the average working white woman may have begun to increase between 1984 and 1986. The CPS results point to the possibility that the long term trend of increasing black women's relative wages due to declining discrimination against black women and improved labor market skills of black women may have ended at precisely the time that white women were beginning to feel the benefits of declining sex discrimination and improved labor market skills. Because these observations are based on a very small change in numbers between the 1985 and 1987 CPS, however, it is important to see if this result persists over several years before accepting that black women's wage growth has indeed stagnated. It should be noted that the CPS results do not coincide perfectly with the results obtained using the 1980 census and the Survey of Income and Program Participation (SIPP). For instance, the CPS data suggest that the average black woman earned 96 percent as much as the average white woman in 1980, whereas the 1980 census suggests a figure of 94 percent. Similarly, the CPS suggests a wage ratio of 96 percent for 1984, whereas the SIPP yields a figure of 91 percent. Moreover, the 1984 CPS finds a much smaller unexplained wage differential than the SIPP. Given the degree of variation among these three data sets, it is not possible come to firm conclusions about the precise level of the black-white female wage ratio. Similarly, it is not possible to pinpoint precisely the degree to which black women earn less than white women with the same characteristics. Yet, broad conclusions are possible: all three data sets agree that, depending on the characteristics considered, black women today earn between 90 percent and 100 percent as much as comparable white women. Compared to the ratios of 50-66 percent found for earlier years, this is a relatively small range of variation. The CPS results presented in this section have served two purposes. First, they have provided evidence concerning trends in black- white wage differentials over the 1980s. They suggest that the long term trend of increasing relative wages and decreasing discrimination for black women may have stopped, possibly even reversed. Second, because the CPS results overlap with both the 1980 census and the SIPP, they allow assessment of the overall reliability of these results. Given the degree of disagreement of different data sets covering the same years, it appears that it is possible to provide only a broad picture of black women's labor market status. The general trends appear to be correct, but no single number can be accepted uncritically. ENDNOTES (1) Labor market discrimination may lower black women's skill indirectly, however. If black women are paid less for their skills or if they are barred from using their skills at all because of labor market discrimination, they will have a lesser incentive to acquire them. (2) Examples of previous research on the wage effects of labor force experience, job tenure, and intermittent work include Jacob Mincer and Solomon Polachek, "Family Investments in Human Capital and the Eranings of Women," Journal of Political Economy, vol. 83, no. 2 (1974), pp.S76-108; Steven Sandell and David Shaprio, "An Exchange: the Theory of Human Capital and the Earnings of Women," Journal of Human Resources, vol. 13, no. 1 (1978), pp. 103-17; and Jacob Mincer and Haim Ofek, "Interrupted Work Career: Depreciation and Restoration of Human Capital," Journal of Human Resources, vol. 7, no. 1 (1982), pp. 3-24. (3) It should be noted that years of work experience do not necessarily reflect labor market skills accurately. If black women are denied access to on-the-job training, then their skill levels may not be commensurate with their years of work experience. Moreover, wages may incerase with years of work experience because of seniority systems rather than because of increased skill levels. (4) The statistical procedure adopted here was developed by James J. Heckman, "Sample Bias as a Sepcification Error," Econometrica, vol 47 (1979), pp. 153-62. (5) These results using SIPP data are contray to the results obtained using 1980 census data, which can imply that racial differences in characteristics can account for the entire wage differential and that no differences in pay structures remain. (6) Women over 40 years old at the time of the SIPP (in 1984) would have been over 20 years old in 1964. Thus, they received all of their education and entered the labor makret before the major thrust of school integration in the 1960s and 1970s and before the passage of the 1964 Civil Rights Act. (7) Estimating separate regressions by region and by age group confirms that this result holds true both for younger and for older black women in the South. However, the occupational disparities between black and white women in the South are greater for older women. (8) Since older women are more heavily represented among black women with low educational levels, it seems reasonable to suppose that the reason poorly educated black women are in lower paying occupations than similarly educated white women is that they are predominantly older women who were educated and began working before 1960. Estimating separate regressions by age group and education does not confirm this hypothesis. Younger black women with less than a 12th grade education are also found to be in lower paying occupations than equally qualified white women. (9) These results are contrary to the results obtained in chap. 8 using census data, which suggested that highly educated black women and black women outside of the South earned more than comparable white women. (10) The same approach was adopted in chap.7 for the Census of Population data. (11) Regression results for the MArch 1980 CPS data are reported in app. F, tables F.3 and F.4. Results for the March 1985 data are reported in tables F.5 and F.6, and results for the March 1987 data are reported in tables F.7 and F.8.