Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
q Occupational Segregation and ~ Labor Market Discr~nabon FRANCINE D. BLAU The post-WorIcl War II period has wit- nessed a rapid growth in female labor force participation and a steady narrowing of sex differences in the extent of participation in work outside the home. In 1950, 86.8 per- cent of the (adult) male population partici- pated in the labor force as compared with 33.9 percent of the (adult) female population (U.S. Dept. of Labor, ETA, 19811. By March 1982, 76.6 percent of males and 52.1 per- cent of females were labor force partici- pants.~ The large increase in participation ~ Black women's participation rates have historically been considerably higher than those of white women, although the differential has declined in recent years. In 1955 the labor force participation rate of black and other nonwhite women was 46.1 percent in comparison with 34.5 percent for whites. By 1980 the participation rates were 53.4 and 51.3 percent for nonwhites and whites, respectively. The participation rates of black males have fallen more sharply than those of white males over this period. In 1955 the participation rate of both white and nonwhite males was about 85 per- cent; by 1980 the participation rates of whites and non- whites were 78.3 and 70.8 percent, respectively (U.S. Dept. of Labor, BLS, 1980, 1981~. These differing ra- cial trends in participation rates are an important factor in evaluting the recent gains in black earnings noted below. See Brown (1981~. ~7 rates of married women (with husband pre- sent) from 21.6 percent in 1950 to 51.0 per- cent in March 1981 (U.S. Dept. of Labor, BLS, 1982) was a major factor in the expan- sion of the female labor force. These trends appear to have been accom- panied by an increase in the labor force at- tachment of women. This is suggested by the marked rise in the labor force partici- pation rates of married women (with hus- band present) with preschool-age children from 11.9 percent in 1950 to 47.8 percent in March 1981 (U.S. Dept. of Labor, BLS, November 19811. Further evidence of the increasing stability of women's participation is provided by the decrease in labor force turnover among women that has occurred over the last 20 years, particularly since the late 1960s (Lloyd and Niemi, 19791.2 At the same time, the gap between men's and women's educational attainment (average 2 Labor force turnover is measured by the ratio of the labor force experience rate (the percentage of women who were in the labor force at some time during the year) to the annual average labor force participation rate (the percentage of women who where in the labor force at any particular point in time or survey week).
118 . . FRANCINE D. BLAU years of schooling) has been eliminated (Lloyd and Niemi, 1979~. These dramatic shifts in the quantity and quality of labor supplied to the market by women do not appear to have been accom- panied by a noticeable improvement in the relative economic status of women workers. In fact, the median earnings of year-round, full-time women workers as a percentage of men's fell from 63.9 to 60.2 between 1955 and 1981 (Lloyd and Niemi, 1979; U. S. Dept. Of Commerce, 1982~. Most of the decline had occurred by the early 1960s, and the earnings ratio has been roughly stable since then. The occupational distributions of men and women continue to differ significantly. A small movement toward greater similarity appears to have occurred between 1960 and 1970 based on census data (Blau and Hen- dricks, 1979~. Some estimates suggest that the rate of decline in segregation may have accelerated between 1972 and 1981, al- though the magnitude of segregation re- mained substantial (Belier, this volume).3 Throughout this period over 60 percent of the female (or male) labor force would have had to change jobs to eliminate the over- representation of women in certain occu- pations and their corresponding underre- presentation in others (Blau and Hendricks, 1979; Beller, this volume). The precise role of labor market discrim- ination in producing these sex differentials in occupational distributions and earnings has been the subject of considerable debate among economists. In the first section we consider the explanations that have been of- 3 Lloyd and Niemi (1979) find no change in the de- gree of segregation over this period using census data for 1970 and Current Population Survey (CPS) data for 1977. However, using comparable CPS data, Beller (in this volume) finds that segregation continued to decline over the 1970s. Full resolution of this disparity in re- sults may await the availability of the 1980 census data. In any case both studies suggest that the magnitude of segregation remained quite high. fered for sex differentials in employment patterns and earnings. We focus on the var- ious neoclassical and institutional models of sex (and race) discrimination and on the al- ternative explanation provided by the hu- man capital model, which emphasizes the voluntary choices of women. In the second section we first evaluate the empirical evi- dence on the degree of sex discrimination in the labor market, and then turn to an assessment of the role of employment seg- regation by sex in producing differential out- comes for men and women workers. ECONOMIC EXPLANATIONS OF SEX DIFFERENTIALS IN OUTCOMES Theories of Discrimination While most of the discrimination models discussed here were developed to explain or at least were illustrated in terms of- racial differences, we here apply them to sex differences. The foundation for the mod- ern neoclassical analysis of labor market dis- crimination was laid by Becker (19571. For simplicity, it is assumed that male labor and female labor are perfect substitutes. That is, men and women are equally productive and thus deserving of equal wages in the absence ofdiscrimination. Discriminatory tastes may exist in employers, coworkers, and/or cus- tomers. Employers with "tastes for discrimina- tion" against women will hire women work- ers only at a wage discount that is sufficiently large to compensate them for the disutility of employing women. Becker also showed that even if employers themselves have no tastes for discrimination against women, profit-maximizing behavior by employers may result in sex discrimination if employees or customers have such discriminatory tastes. Male employees with tastes for discrimi- nation against women will work with them only at a wage premium that is sufficient to compensate them for the disutility offemale
OCCUPATIONAL SEGREGATION AND LABOR MARKET DISCRIMINATION 119 coworkers.4 Customers with tastes for dis- crimination against women will buy prod- ucts or services produced or sold by women only at a lower price. Intuitively we would expect this type of discrimination to be more important in sales or service occupations where face-to-face contact with the customer/client occurs. As a consequence of coworker or cus- tomer discrimination employers may, under certain circumstances, discount female wages to compensate for the higher costs (coworker discrimination) or lower revenues (customer discrimination) attendant upon employing women. A definition of wage discrimination flow- ing from Becker's work has guided much of the empirical analysis of labor market dis- crimination. Wage discrimination (the mar- ket discrimination coefficient) may be de- fined as the difference between the actual ratio of male to female wages and the ratio that would exist in the absence of discrim- ination assuming perfect substitutability, this would be wage parity (Becker, 1957, p. 1261. In empirical work, where the wages of heterogeneous male and female labor are compared, this is approximated by the no- tion of pay differentials that are not ac- counted for by productivity differentials. While the type of discrimination defined by Becker does not necessarily predict that occupational segregation by sex will occur, it may be made compatible with occupa- tional segregation if we postulate that tastes for discrimination against women vary across occupational categories. The issue may be more one of socially appropriate roles than of the desire to maintain social distance that Becker emphasized. Employers may have no compunctions about hiring women as secretaries but may be reluctant to employ 4 The notion of employee discrimination is developed by Bergmann and Darity (1981) in terms of productivity reductions due to employee hostility rather than of direct increases in costs due to compensating differ- entials. them as pipefitters. Men may be willing to work with women in complementary (co- operative) or subordinate positions but dis- like interacting with women as peers or superiors. Customers may be delighted to purchase nylons from female clerks but avoid women car salespersons or attorneys. These discriminatory tastes may be held independ- ently of beliefs that women would be less pro- ductive than men in nontraditional pursuits. Ibis latter possibility is considered under no- tions of statistical discrimination below. While such reasoning makes Becker's mode} more compatible with the large mag- nitude of occupational segregation that we observe in the labor market, problems re- main. First, as Blau and Jusenius (1976) point out, a high degree of segregation is stfl} un- likely given (1) the wage flexibility generally assumed in neoclassical models and (2) the existence of a large ceteris Paris wage dif- ferential between men and women. (The empirical work considered below provides support for the existence of a substantial pay gap that is not accounted for by the pro- ductivity-related characteristics of men and women.) For example, let us consider the case in which discriminatory tastes reside in employers. Employers whose tastes for dis- crimination are so strong that they exceed the marketwide discrimination coefficient wit} not hire women. Employers who are exactly compensated for the disutility of hir- ing women by the market discrimination coefficient will be indifferent to whether they employ men or women and will presumably hire both. Employers with relatively weak tastes for discrimination who are over- compensated by the marketwide discrimi- nation coefficient will hire only women. A high degree of occupational segregation by sex due to discrimination will not be ob- served unless most employers of workers in "male" jobs are in the first category, that is, unless most employers have such strong tastes for discrimination against women in male jobs that they are not sufficiently compen- sated by the large ceteris paribus pay dif-
120 FRANCINE D. BLAU ferentials that appear to exist between men and women. This seems to be unlikely. Differences in tastes for discrimination among employers (employees, customers) can perhaps more plausibly produce sex segre- gation by firm in the Becker mode} (Arrow, 19731. That is, women would tend to be em- ployed by less discriminatory employers who are overcompensated by the prevailing sex pay differential. A problem that arises here is the stability of this situation in the face of competitive forces (see below). Women may also, in this model, find better employment opportunities working with less discrimina- tory employees or selling to (serving) less clis- criminatory customers (clients). Second, in the Becker mocle! cliscrimi- natory pay differentials are in some sense the price paid by the discriminated group for associating with the discriminators. In general, differences between the two groups in factor endowments make such association profitable even in the face of discrimination. However, sufficient opportunities in a seg- regated context can eliminate the need for pay differentials. This may be illustrated by the case in which tastes for discrimination reside in coworkers. In the case of perfect substitutes, for example among workers in the same occupational category, complete sex segregation by firm is expected, since sexually integrated work forces are more costly (i.e., men must be paid a premium to work with women). The necessity for wage differences is obviated by such segregation, however, since men and women do not work together (and thus it is not necessary to com- pensate male workers for the disutility of working with women). Discriminatory pay differentials will arise in this case only if for some reason (e.g., costs of adjustment due to personnel investments in workers fArrow, . 1973] ~ complete segregation is not possible.5 5 Of course, one may wonder why men rather than women are the recipients of these personnel invest- ments. As Arrow (1973) points out, where there are From this perspective one may question why a high degree of occupational segregation by sex appears to be associated with large dis- criminatory pay clifferentials. Even if discrimination is made compatible with occupational segregation in the Becker model, segregation does not play a causal role in generating the sex pay differential. Rather, both occupational and pay differ- entials are due to tastes for discrimination among employers, coworkers, and/or cus- tomers. Eliminating occupational segrega- tion (were it possible to do so) would not eliminate the pay differential. Indeed, re- ducing segregation might require still fur- ther discounting of female wages, since it would necessitate women's making inroads into areas characterized by stronger discrim- inatory tastes on the part of employers, co- workers, and/or customers. Bergmann (1974) has developed an anal- ysis in the Becker tradition that gives a more central role to employment segregation. In Bergmann's "overcrowding" model, 6 dis- criminatory employer tastes result in the segregation of male and female labor into two sets of occupation. While such segre- gation need not result in sex pay differen- tials, differentials will occur if job oppor- tunities (clemand) in the female sector are small relative to the supply of female labor. Employers who do hire women will utilize the labor-intensive production techniques that their lower wages make profitable. Thus, costs to change history matters. Given historically ris- ing female participation rates, women, as relatively new entrants, may find males already in place in many sec- tors. However, women have always been heavily con- centrated in a few female-dominated activities, even when they constituted a small proportion of the labor force. Assuming discrimination played a role in this segregation, it appears that the notion of the personnel investment tends to require both employer and em- ployee discrimination. Further, once we have person- nel investments men and women are no longer in fact equally productive, although they may be potentially equally productive (see below). 6 See also Edgeworth (1922) and Fawcett (1918).
OCCUPATIONAL SEGREGATION AND LABOR MARKET DISCRIMINATION in contrast to Becker's (1957) analysis, seg- regation may play a causal role in producing discriminatory pay differentials. Further, discrimination may cause both pay and pro- ductivity differentials between potentially equally productive male and female labor women are less productive than men be- cause, as a result of segregation and crowd- ing, they have less capital to work with. The Bergmann formulation does not overcome the problem with the Becker model, noted earlier, that an extreme distribution of em- ployer tastes is necessary to generate the high level of segregation we observe. How- ever, while Bergmann postulated employer discrimination as the source of segregation, the overcrowding concept may be linked to any postulated reason for segregation. It is thus a persuasive explication of the wage consequences of segregation, regardless of its cause. Another question that has been raised about the Becker analysis, particularly with regard to the case of employer discrimina- tion, is the issue of the survival of discrim- ination in the long run under perfect com- petition (Arrow, 19731. Assuming that employer tastes for discrimination against women vary, the least discriminatory firms that hire the highest proportion of (Iower- priced) female labor will have lower costs and thus higher profits. Capital will flow toward these firms, and, assuming constant returns to scale, only the least discrimina- tory (Iowest-cost) firms will survive. The ap- parent persistence of sex (and race) discrim- ination in the labor market over time has given rise to additional analyses of discrim- ination, which we shall consider below. However, this criticism of the Becker mode} is a double-edged sword in that it creates skepticism among many economists that la- bor market discrimination is indeed respon- sible (in whole or part) for the observed sex clifferences,in 'market outcomes. Perhaps the best d~oped alternative explanation is the human capital mode} considered in the next section. 121 One obvious solution to the problem raised above is that noncompetitive elements are responsible for the persistence of discrimi- nation. Becker (1957) hypothesized that em- ployer discrimination should, on average, be less in competitive industries than in mo- nopolistic ones. In the case of sex discrim- ination the focus has tended to be on im- perfections in the factor market rather than in the product market. Madden (1973) has developed Robinson's (1933) monopsony mode} to explain sex differences in wages. Monopsony describes the situation in which labor faces a single buyer. A price-discrim- inating monopsonist will pay female labor less if it is less elastically supplied to the firm than is male labor. Assuming that the supply curve of labor to the firm is positively sloped, the elasticity of labor supply to the firm is the percentage increase (decrease) in labor hours supplied to the firm in response to a given percentage increase (decrease) in the wage offered by the firm. A lower elas- ticity of labor supply for women thus means that the quantity of labor supplied by women to the firm is less responsive to wage changes than is the case for men. The persuasiveness of this explanation for aggregate pay differentials by sex is unclear, a priori. One issue relates to Madden's (1973) argument that female labor is less elastically supplied to the firm. On the one hand, as Madden (1973, 1976) argues, such factors as occupational segregation and the power of male unions may limit women's alternatives and thus decrease their wage elasticity of supply to the firm, all else equal. Supply- side factors, such as the tendency for women to engage in less job search than men do or to seek jobs that are closer to home, could also contribute to this result. On the other hand, as Blau and Jusenius (1976) argue, the aggregate female labor supply curve (to the market) is more elastic than the male labor supply curve (to the market). This has con- sequences for the elasticity of supply to the firm in that home work provides a viable alternative for women at the margin of labor
122 FRANCINE D. BLAU - force participation.7 Furthermore, if men are more likely than women to acquire firm- specif~c training, that would also lower their mobility relative to women. A second issue relates to the degree of monopsony that ac- tually exists in the labor market. Pure mon- opsony (one buyer) in a labor market is un- doubtedly quite rare. However, Madden (1973) argues that there are considerable monopsonistic elements in the labor mar- ket. This is an empirical issue deserving of more attention. Moreover, it is not clear that the case of few buyers of labor can be analyzed in the same way as can the case of one buyer. It seems likely that the monopsony ex- planation is more applicable to specific oc- cupations and labor markets than to the ag- gregate sex pay differential.8 An ideal case might be the employee! female (or male) Ph. D. with an employed Ph. D. spouse in a one-university town. Nonetheless, this the- ory does set up a mechanism by which oc- cupational segregation may lower women's wages relative to men'sin this case by reducing women's options. A second approach to explaining the long- run existence of discrimination is the notion of statistical discrimination (Phelps, 1972; Arrow, 1972a,b, 1973; Aigner and Cain, 19771. Statistical discrimination provides a motivation for discrimination that is com- patible with profit-maximizing behavior on the part of employers. It stems from im- perfect information and may take one of two forms. 7 Women are more likely than men are to quit their jobs to leave the labor force, while men are more likely than women to quit to change jobs (Barnes and Jones, 1974~. 8 Some preliminary evidence in support of the mon- opsony view as an explanation for male-female wage differentials across urban areas, based on data for white males and white, never-married females, is presented by Cardwell and Rosenzweig (1980~. Note that in re- stricting their analysis to never-married females they focus upon women with the lowest value of nonmarket time. First, employers may discriminate against women because of real or perceived average productivity or productivity-related behav- ioral differences between men and women. In this case sex is assumed to provide in- formation regarding expected procluctivity. Aigner and Cain (1977) argue that economic discrimination does not exist if the employ- er's perception of the average sex difference is correctsince on average each group is paid in proportion to its productivity. How- ever, they acknowledge as disquieting the result that at each ability level women will receive lower pay than mend Others have called this discrimination in that the "indi- vidual is judged on the basis of the average characteristics of the group . . . to which he or she belongs rather than upon his or her own characteristics" (lturow, 1975, p. 172~. is Implicit in this view seems to be the as- sumption that other personal characteristics besides sex are readily available and that while the sex difference exists, on average it would not be present in a ceteris paribus comparison. If so, perhaps this might be more appropriately considered mistaken be- havior on the part of employers. But, as Aigner and Cain (1977) point out, discrim- ination based on employers' mistaken be- liefs is as unlikely (or even more unlikely) to persist over time in the face of compet- itive forces as is discrimination based on em- ployer tastes. So the question of the per- sistence of discrimination in the long run remains. 9 This assumes that the variances of the measurement error and of the productivity indicator are the same for males and females. it See also Ellau and Jusenius (1976) and Piore (1971~. Lewin and England (1982) argue that it is the explicit use of ascriptive characteristics like race or sex in per- sonnel decisions that constitutes the discriminatory as- pect of statistical discrimination, even when the em- ployer perceptions are correct. From a normative perspective this is certainly correct, but it is not clear that it counters Aigner and Cain's (1977) argument that such behavior does not constitute economic discrimi- nation.
OCCUPATIONAL SEGREGATION AND LABOR MARKET DISCRIMINATION 123 Arrow's (1973, 1976) notion of perceptual equilibrium sheds some light on this issue. Arrow's model allows for the endogeneity of qualifications workers become qualified by making some type of investment in them- seIves where the decision to invest depends on the gain from qualifying. In this case em- ployers' perceptions of sex differences in qualifications may become self-confirmin~ even when there is no intrinsic sex differ- ence in ability or behavior. Multiple equi- libria may result. For example, if employers' view of female job instability leads them to give women less training and to assign them to jobs where the cost of turnover is mini- mized, women may respond by exhibiting the unstable behavior employers expect. This in turn confirms employer perceptions. On the other hand, if employers believe women are stable workers, they will hire women into positions that are sufficiently rewarding to inhibit instability (Arrow, 1976. Here, pay and productivity differences between potentially equally productive male and fe- male workers that may persist in the long run can result from employers' in some sense erroneous beliefs. Viewing the matter some- what differently, the employers' ex post "correct" assessment of sex differences in average productivity may be seen to result from their own discriminatory actions. Moreover, the resulting female sector may be subject to a Bergmann-type overcrowd- ing problem, further reducing relative wages there. A second type of statistical discrimination may occur even if the two sexes have equal average abilities or behavior. Risk-averse employers may discriminate against women if their ability or behavior is less reliably predicted by some indicatorks) than is men's (Phelps, 1972; Aigner and Cain, 19771. How- ever, Aigner and Cain express doubts that risk aversion could cause discriminatory pay differentials of the magnitude obtained by ~i See also Spence (1973, 1974~. empirical studies (see below). As they note, dispersion in risk aversion among employers should result in the bidcling up of women's wages, just as the existence of less discrim- inatory firms should erocle discriminatory pay differentials in the Becker-type taste- for-discrimination model. Further, the ex- istence of a large risk discount of women's wages should trigger a market for test in- struments or indicators that are equally re- liable for them. At present neither the role of occupa- tional segregation nor the issue of the per- sistence of discrimination in the long run appears to be satisfactorily understood at the theoretical level. i2 Perhaps the most prom- ising notion advanced here is Arrow's idea of perceptual equilibrium, a kind of "vicious circle," or feedback theory of discrimina- tion, which, as noted earlier, can accom- modate and provide some rationale for oc- cupational segregation. i3 The overcrowding concept can be appended to this model to imply a further reduction in wages due to segregation. A problem, however, is that this mode} cannot explain the sex segrega- tion that appears to exist among jobs re- quiring similar amounts of skill, stability, etc. While this may not be a major com- ponent of occupational segregation, it is probably a nontrivial component. Institutional models, such as the internal labor market analysis or the dual labor mar- ket mode! (Doeringer and Piore, 1971; Piore, 1971),~4 that give a more explicit role to oc- cupations may be helpful here. Blau and Jusenuis (1976) argue that the major con- tribution of such models is not to suggest i2 The latter point is emphasized in Darity's (1982) consideration of racial pay differentials. i3 A number of other authors have emphasized the importance of feedback effects in analyzing sex pay and occupational differentials. See, e.g., Bergmann (1976), Blau (1977), Ferber and Lowry (1976), Strober (1976), Weiss and Gronau (1981~. t4 See also Cain (1976), Gordon (1972), and Wachter (1974~.
124 FRANCINE D. BLAU new explanations for sex differentials in earnings and employment distributions but rather to elucidate the linkage between the two: to suggest why the same set of factors that produce earnings differentials is also likely to generate employment segregation. They argue that under the administered sys- tem of the internal labor market, the firm attaches wage rates to occupational cate- gories rather than to individuals (see also Thurow, 19751. Under such circumstances the only way in which the firm can distin- guish between men and women in terms of pay is to assign them to different job cate- gories. Within occupational categories, in- stitutional constraints mandate equal pay for equal work, except for relatively small dis- tinctions based on seniority and/or merit considerations. Such group treatment of in- dividuals will be most efficient (wfl} result in the discarding of the least information) if occupational categories are as homogenous as possible. Thus, employers are likely to structure female jobs to fit the perceived av- erage characteristics of women workers. This, in turn, influences women's behavior and pro- ductivity a la Arrow's perceptual equilibrium. Bergmann's overcrowding mechanism may further widen the pay differential between male and female jobs. The Human Capital Alternative Theories of discrimination are concerned with explaining occupational and pay differ- entials between (potentially) equalEy produc- tive men and women. Me aggregate figures cited earlier in fact compare heterogeneous male and female labor. Some or ad of the observed occupation and pay differences may in fact be due to productivity differences be- tween the sexes. The human capital mode! can provide a consistent explanation for oc- cupation and pay differentials by sex in terms of the voluntary choices of women rather than of market discrimination against them. It then becomes an empirical question, albeit a dif- ficult one, to determine which view is correct or what portion of the observed differences is accounted for by each explanation. As developed by Mincer and Polachek (1974) and others, the human capital anal- ysis calls attention to the traditional division of labor by sex within the family under which women can expect shorter and more dis- continuous involvement in market work than can men. This reduces their long-run payoff to human capital investments, since they have a shorter work life over which to reap the returns. Similarly, employers will be re- Juctant to invest in firm-specific training for women workers. Thus, women may earn less than men both because of their lesser amount of labor market experience and because of the lower returns to experience they obtain (the latter reflecting their smaller invest- ments per unit of time). Female earnings are further reduced by the depreciation of their stock of human capital during the time they spend out of the labor force. The human capital mode! can also provide a theory of occupational choice to explain occupational segregation by sex (Polachek, 1976, 1979, 1981; ZelIner, 1975; Landes, 19771. According to this view women wit! tend to avoid occupations requiring consid- erable investments in on-thejob training and having high rates of depreciation for time spent out of the labor force. Thus, we would expect predominantly female jobs to have relatively flat age-earnings profiles. One problem with this analysis is that the sexual division of labor within the family is in turn influenced by the relative market rewards (wage rates) of husband and wife (Becker, 1965, 1973~. This implies that labor market discrimination against women could influence their allocation of time between the home and the market and thus the amount and types of their human capital invest- ments. From the standpoint of empirical work this means that analyses which treat such variables as experience and education as ex- ogenous may underestimate the extent of
OCCUPATIONAL SEGREGATION AND LABOR MARKET DISCRIMINATION 125 labor market discrimination (Weiss and Gronau, 1981~. Further, women's Tower re- turns to experience may reflect employers' reluctance to provide opportunities for firm- specific training as well as their own vol- untary decisions. In the latter case it would also be important to determine whether the employers' decisions are in fact justified by ceteris Paribas sex differences in quit pro- pensities. EMPIRICAL FINDINGS As indicated in the preceding section, there are alternative views as to how labor market discrimination might produce occupational and pay differences between men and women. Further, the human capital mode} provides an alternative explanation for sex differences in market outcomes that is con- sistent with voluntary decision making by women rather than with discrimination against them. For the most part empirical research has been focused on the question of whether or not labor market discrimina- tion against women (and minorities) exists. Choosing among alternative models of dis- crimination and understanding the causes and consequences of employment segrega- tion have received considerably less atten- tion, particularly from economists. In this section we first consider in some detail the question of whether or not dis- . . . . ,.,,. . . . . cr~m~nat~on exists. ~ nits Is an Important ~n- quiry, since its resolution is necessary in order to determine the overall context in which employment segregation by sex takes place. There is no point in seeking to de- termine the role of segregation in producing discriminatory pay differentials by sex if in fact there is little evidence that such dis- criminatory differentials exist. Second, we explore the existing literature concerning the role of occupational segregation in produc- ing pay differentials and, more briefly, some of what has been learned about the causes of such segregation. Discrimination and Earnings A crude test of the relative merits of the discrimination and human capital explana- tions for sex differences in earnings is pro- vided by an examination of the time series trends in the sex pay differential. We first briefly consider this time series evidence, then move to a detailed discussion of the more sophisticated cross-sectional analyses ofthe extent of labor market discrimination. As noted in the introduction, there has ac- tually been some deterioration in the rela- tive earnings position of women since the mid-19SOs. It has frequently and incorrectly been assumed that increases in the female labor force participation rate over time are indicative of declines in the average level of experience of women workers due to the impact of new entrants (Economic Report of to President, 19741. In fact, as Mincer (1979) explains, the female labor force grows not only through "widening" (increases in the flow of entrants or reentrants) but also through "deepening" (decreases in the flow of exi- ters). Ike impact on the average level of experience of female workers depends on the relative magnitude of these two flows. In addition, since labor force entry tends to be selective of female nonparticipants with higher levels of previous labor force expe- rience, and labor force exit tends to be se- lective offemale participants with lower lev- els of previous labor force experience, labor turnover does not dilute average experience levels as much as it might first appear (Blau, 1975, 19781. Lloyd and Niemi (1979) present a variety of evidence indicating that the trend has been for women to remain in the labor force longer and more continuously and that en- trants make up a decreasing proportion of the female labor force. They conclude that "it appears . . . the work experience di~er- ential between the sexes has narrowed in the past twenty years" (p. 133~. Thus, female participation trends do not seem to be re-
126 FRANCINE D. BLAU sponsible for the widening pay gap. More- over, they find that "over time, the gap be- tween men's and women's educational attainment has been eliminated and, among young cohorts, it appears that women's pro- pensity to enroll in college is roughly similar to men's" (p. 1461. Men and women do con- tinue to differ sharply in fields of speciali- zation, although some progress has recently been made in this area as well (Baker, un- dated). Further, one may speculate that the rising divorce rate has increased women s incentives to invest in their labor-market- related human capital. While one would like better data, partic- ularly on experience, it seems reasonable to conclude that human capital factors do not account for the widening pay gap between men and women. This provides support for the view that discrimination plays a role in producing the differential. It is unlikely that tastes for discrimination against women have increases! during this period. However, as we have seen, little decline occurred in the degree of sex segregation in employment during that time. If increases in the demand for labor in female jobs did not keep pace with increases in the supply offemale labor, relative "crowding" in female jobs may have worsened, exerting a downward pressure on female wages relative to male wages. In- creases in the real wages of women over the period could have continued to induce in- creases in female labor force participation celery Paribas. In addition, it may be argued that women face substantial experience and training re- quirements in their efforts to move into high- level male jobs (Freeman, 19731. This makes it difficult for equal employment opportu- nity legislation to open the doors to such jobs for older women. Further, younger women may have had the opportunity to incorporate new expectations of greater la- bor force attachment over their life cycles into their human capital investment deci- sions to a greater extent than have older women. It is true that younger women have been increasing their earnings position rel- ative to younger men. Between 1970 and 1981 the ratio of the median incomes of full- time, year-round women workers to those of men increased from 59 to 60 percent for all women, but rose from 65 to 70 percent among 25- to 34-year-olds (U.S. Dept. of Commerce, 1980, 19821. Beller (this vol- ume) also found that younger cohorts ex- perienced greater declines in occupational segregation than clid the work force as a whole. Thus far, however, the modest gains by younger women do not provide over- whelming support for this view. Much will depend on how this group fares in the next few years. It is also important to point out that if employers were willing to reevaluate their traditional promotion ladders they might find that many older women workers do have the experience necessary for higher-level positions. During recent years, while women as a group have ma(le little progress in advanc- ing their earnings positions relative to men, black women have advanced relative to white women. Black women's median (full-time, year-round) incomes have increased from 69 percent of those of white women in 1964 to 90 percent in 1981. Black men have also gained relative to white men. Their median full-time, year-round incomes rose from 66 percent of those of white men in 1964 to 71 percent in 1981 (Lloyd and Niemi, 1979; U. S. Dept. of Commerce, 19821. The gains in the relative incomes of blacks are partially due to substantial increases in their relative educational attainment but cannot be fully explained by that factor (Brown, 1982~. They may in part be due to the impact of equal employment opportunity legislation. The more rapid improvement in the relative in- come position of black women (compared with white women) than of black men (com- parecl with white men) may reflect the large number of entry-level positions in many typically female jobs, in comparison with the higher experience and training require- ments in typically male jobs discussed above.
OCCUPATIONAL SEGREGATION ._ - LABOR MARKET DISCRIMINATION 127 We now turn to a review of studies that attempt to measure the extent of labor mar- ket discrimination explicitly. The general practice in empirically estimating the mag- nitude of labor market discrimination against a particular group, e.g., women, is to as- certain the proportion of the sex differential that is accounted for by differences in the productivity-related characteristics of male and female workers and to allocate the re- sidual to discrimination. Discrimination may be measured by the coefficient on a sex dummy variable in an earnings regression equation, in which case the impact of the other explanatory variables on earnings is constrained to be the same for each sex group. More often, this constraint is relaxed by es- timating separate earnings regressions for sex or race-sex groups. In this case the por- tion of the pay differential due to sex clif- ferences in the returns to a given set of char- acteristics (i.e., sex differences in coefficients, including the constant term) is attributed to discrimination; the portion of the pay dif- ferential that is due to sex differences in endowments of productivity-related char- acteristics (i.e., sex differences in the means of the inclependent variables) is attributed to nondiscriminatory factors (Blinder, 19731. There are various problems with this "re- sidual" approach to measuring labor market discrimination. Perhaps the most serious is the specification problem. On the one hand, conventional data sources do not allow for the measurement of all productivity-related characteristics. The absence of actual labor market experience from the data sets uses! in the early studies of sex discrimination is a prime example of an important omitted variable. Type (as opposed to amount) of schooling, which varies greatly between men ant! women, would be another example. If, on average, males are more favorably en- clowed with the characteristics measured by these omitted variables, the extent of labor market discrimination will be overestimated because of imperfect controls for these omit- ted factors. On the other hancI, group dif- ferences with respect to some productivity- related characteristics may reflect the indi- rect effects of discrimination (Blinder, 19731. For example, as discussed earlier, women may be excluded from high-wage occupations be- cause of their sex. Furler, if the endogeneity of choice variables is taken into account, we see that labor market discrimination may dis- courage women's human capital accumulation or alter its type. Measured labor market clis- crimination is likely to be underestimated to the extent that such factors representing other dimensions of discrimination are controlled for. A related problem is the interpretation of sex differences in the coefficients of earnings regressions. For example, a smaller coeffi- cient on labor force experience for women may reflect their decisions to invest in less on-thejob training than men do, as pro- posed by human capital theorists, or dis- crimination on the part of employers re- sulting in less access to on-thejob training opportunities. 15 Similarly, sex differences in the returns related to marital status may re- flect unmeasured differences in labor force attachment between married men and women (Polachek, 1975) or sex differences in employer treatment of marital status that is unrelated to productivity (e.g., the view that married men deserve higher salaries because they have families to support). These problems of specification and in- terpretation of coefficients reflect a more fundamental problem. We would like to measure the extent of labor market dis- crimination a demand-side phenomenon. But wages are influenced by both sup- ply- and demand-side factors. The earnings functions that are typically estimated are es- sentially reduced-form equations, and thus their coefficients may reflect the influence i5 Further, one may question the basic premise of the human capital model that upward-sloping experi- ence-earnings profiles are indeed due to on-thejob training. See, e.g., Medoff and Abraham (1980, 1981~.
28 FRANCINE D. BLAU of both supply and demand (Butler, 1981; Chiplin, 19811. Ideally, one would like to specify and estimate a structural mode} with separate supply and demand equations. No one has yet attempted to specify and esti- mate such a mode! on an economyw~cle ba- SiS. An additional problem with the earnings regression approach is that, while the theory specifies the measurement of discrimination in teas of pay differences that are not ac- counted for by productivity differences, the empirical work involves adjustment using proxies like education and experience for productivity. This gives rise to an errors-in- variable problem (Hashimoto and Kochin, 1980; Roberts, 19801. For simplicity, let us assume that the productivity proxies are measured so that they are positively related to earnings. Then it may be shown that if women have lower mean values of the pro- ductivity proxies than men have (as is gen- erally the case), the coefficient on a sex dummy variable (female= 1) in an earnings regression is biased downward, giving an exaggerated estimate of the effect of dis- crimination (Hashimoto and Kochin, 19801. Unfortunately, however, there is no ob- vious solution to this problem. Roberts (1980) suggests a procedure that he terms "reverse regression," in which the independent var- iable (e.g., education) is regressed on the dependent variable (e.g., earnings) and a sex dummy. Such a procedure will pro- duce unbiased coefficients only if the de- pendent variable is measured without error. But earnings as measured are only an im- perfect indicator of permanent earnings ca- pacity, which, one might argue, is the the- oretically relevant variable (Hashimoto and Kochin, 19801. Moreover, measured earn- ings are only a proxy for the total rewards for the job, including fringes and the non- pecuniary benefits of the work (Madden, 19821. Hashimoto and Kochin (1980) suggest ~6 See also Kamalich and Polachek (1982~. performing the analysis on grouped data where the grouping criterion is independent of the measurement errors. But it may not ~ A. ~ be easy to meet this requirement (or to know whether or not one has met it), and, as they acknowledge, the results can be quite sen- sitive to one's choice of a criterion. Another classical approach to the errors-in-variable problem involves the use of instrumental variables. However, the specification of the appropriate instruments is difficult given the current state of theory and the availability of data (Kamalich and Polachek, 19821. An additional problem in using conven- tionally estimated earnings regressions to measure discrimination most likely pro- duces biases in the opposite direction, that is, leads us to underestimate discrimination. Regressions are generally estimated on the selected sample of labor force participants. But Gronau (1974) has argued that it is the wage offers, not the actual wages of males and females, that should be compared. The distribution of actual wages represents only that part of the offer distribution that is ac- ceptable to job seekers.~7 Thus, according to Gronau, mean female wage offers will be overestimated by restriction of the sample to labor force participants. He provides empirical evidence in support of this con- tention. Since male participation rates (par- ticularly in the prime work ages) are still substantially higher than women's, it may be argued that data on men's wages are con- siderably less affected by selectivity bias.~9 i7 See also Heckman (1974~. ~8 Gronau has argued elsewhere (1973) that the value of time of housewives may be either higher or lower than the market wage of comparable employed women. If the former is true, it is not necessarily the case that restriction of the sample to labor force participants ov- erstates women's wage offers. However, the empirical evidence he presents supports Gronau's (1974) conten- tion. See also Cogan (1980~. i9 As noted earlier, black male participation rates are lower than those of whites. Thus, the same type of selectivity problem discussed in the text affects race comparisons among males.
OCCUPATIONAL SEGREGATION AND LABOR MARKET DISCRIMINATION - 129 Thus, observed wage differences between men and women will underestimate the "true" male-female wage-offer differential. On the other hancI, over time male partic- ipation rates have been declining while fe- male rates have been increasing (see above). Thus, we may be sweeping out more of the female offer distribution and less of the male offer distribution than was formerly done. This may be a partial explanation for our failure to observe a decline in the male-fe- male pay gap over time. Bearing these problems in mind, we con- sider the empirical work on sex discrimi- nation. As noted earlier, most empirical work in this area has focused on the question of whether or not labor market discrimination exists and has attempted to estimate its mag- nitude. While there are still various unre- solved problems in estimating the extent of labor market discrimination, it is not clear from our consideration of them that they result, on net, in overestimates or under- estimates of discriminatory pay differentials. If the evidence suggests that there is labor market discrimination against women, then this will provide some motivation for con- sidering the role played by sex segregation in employment in producing these discrim- inatory outcomes. Not surprisingly, the estimate of the sex pay differential that is due to discrimination varies considerably depending on the group studied, data set employed, and variables controlled for. We shall focus our discussion on studies using national samples of indi- viduals across a variety of occupations.20 For the most part, the earlier studies of male- femaTe pay differentials attributed a sub- stantial portion of the sex pay <differential to discrimination. For example, using 1960 20 For more detailed reviews of the empirical liter- ature, including occupation-specific studies, see Kohen (1975), Lloyd and Niemi (1979), and Treiman and Hart- mann (1981~. 2] The one exception was a study by S=bom (19~; see below. census data, Fuchs (1971) found that sex dif- ferences in individual characteristics could account for only 3 to 15 percent of the dif- ferential. Similarly, Oaxaca (1973a,b), using data from the 1967 Survey of Economic Op- portunity, found that 80 percent of the pay differential between white men and white women and 94 percent of the differential between black men and black women could not be explained by productivity-related in- dividual characteristics.22 As these authors were aware, their lack of data on actual labor force experience cre- ated an important omitted-variable prob- lem.23 The general procedure of estimating experience as the years elapsed since school completion, while fairly accurate for males, is much more questionable for females. In addition to their theoretical contributions noted earlier, Mincer and Polachek (1974) were the first to provide empirical estimates of the impact of labor force experience and time spent out of the labor force on the earn- ings of women using newly available lon- gitudinal data. They analyzed retrospective work history data from the National Lon- gitudinal Survey (NLS) of mature women aged 30 to 44 in 1967. Mincer and Polachek were able to account for 45 percent of the pay gap between white married men and women in that age group on the basis of human capital variables, including actual la- bor market experience and time spent out of the labor force. In arriving at this esti- mate, they were aware of the joint deter- mination of earnings and experience and at- 22 Oaxaca's results including controls for occupational and industrial characteristics are considered below. See also Gwartney and Stroup (1973), Sawhill (1973), and Blinder (1973~. On the other hand, Darity and Myers (1980) found that while the structural equations for wages for white males and females were significantly different, they were not so in the case of black males and females. 23 For an exception, see Suter and Miller (1973), who restricted their sample to women from the National Longitudinal Surveys in the 30- to 44-year age group who had worked continuously since school completion.
130 FRANClNE D. BLAU tempted to adjust for simultaneous equations bias by using a two-stage procedure. While Mincer and Polachek are to be commended for their awareness of the simultaneity prob- lem, their application of the two-stage pro- cedure in their exploratory study is far from thorough. For one thing, their specification of the determinants of lifetime labor supply (the proportion of years the respondent worked 6 months or more since school com- pletion) includes an endogenous variable, number of children. For another, while life- time experience is treated as enclogenous, home time and job tenure (seniority) are entered into the earnings function as ex- ogenous variables (Sandell and Shapiro, 1978; Mincer and Polachek, 19784. That these problems are serious is sug- gested by their counterintuitive findings for the two-stage procedure. In the single-pe- rioc! context, economic theory would lead one to expect a positive relationship be- tween labor force participation ancI own wage. An increase in the market wage avafl- able to a woman increases the opportunity cost of nonmarket activities for her. This encourages her to substitute market work for time spent on housework and leisure. However, in a multiperiod context this pos- itive substitution effect could be outweighed by a negative income effect. The negative income effect arises from the fact that the increase in the wage obtained while the woman is working is like an increase in in- come. At higher-income levels she is ex- pected to demand] more of all goods from which she derives satisfaction, including lei- sure. She may thus consume more leisure over the life cycle by supplying less time to the market over the life cycle. Note that she must work some of the time for the income effect to come into play. The evidence sug- gests, however, that for married women's participation decisions the substitution ef- fect dominates the income effect (where the latter is indicated by the response to changes in husband's income [Mincer, 19621~. Thus, women's labor supply is expected to be pos- itively related to their wages. This reasoning implies that the positive coefficient on experience in an ordinary least squares (OLS) earnings regression is biased upward. This is because some of the esti- mated positive impact of labor market ex- perience on wages is really due to a positive effect of wages on experience (at higher wages, women supply more time to the mar- ket). Similarly, the negative coefficient on home time is biased upward in absolute value. This is because some of the estimated neg- ative impact of home time on wages is really due to a negative effect of wages on home time (at higher wages, women spend less time out of the labor market). Correction for the simultaneous equations bias should thus reduce the size of both coefficients in ab- solute value. On the contrary, Mincer and Polachek (1974) find that, if anything, rees- timation of the earnings function using two- stage least squares yields "larger positive coefficients for (total) experience and stronger negative coefficients for home time" (p. Sad. Further, in estimating the proportion of the pay gap explained by the human capital variables, the actual mean levels of the var- iables are employed. Yet the heart of the endogeneity problem is that wage discrim- ination may have influenced the amount of human capital that women have accumu- lated. Thus, it is likely that Mincer and Polachek and analyses modeled on theirs overestimate the impact of the human cap- ital variables on the sex pay differential. In light of this potential bias it is notable that Mincer and Polachek found that over half of the pay gap between white married men and women could not be explained by the hu- man capital variables and was potentially due to discrimination. Another issue that has been raised re- garding Mincer and Polachek's (1974) find- ings is the generalizability of their results for the 30- to 44-year-old age group to the whole female population as is necessary to draw inferences for the aggregate male- female pay gap. The work of Corcoran (1978, 1979) using a full age range from the 1976 Pane] Study of Income Dynamics (PSID),
OCCUPATIONAL SEGREGATION AND LABOR MARKET DISCRIMINATION _ 131 suggests that their findings may not in fact be generalizable. She found that the wanes of women aged 30 to 44 are much more strongly affected by labor force withdrawals than are those of the broader age range (Cor- coran, 19791. Women in this 30-to 44-year age group are more likely than are women in general to have recently reentered the labor market after a prolonged period of nonparticipation. Corcoran suggests that her findings are consistent with the notion that women's wages are temporarily depressed by labor force withdrawals because of mis- information about job opportunities and/or about their own value as workers.24 In addition to providing work histories, the PSID data provide measures of labor force attachment (i.e., absenteeism due to own flIness; absenteeism due to illness of others; self-imposed restrictions on work hours and/or job location; voluntary part- time work). For an unrestricted age group of women, Corcoran (1978) found that ad- justments for schooling, work history, and labor force attachment accounted for 36 per- cent of the wage gap between white men and white women and 27 percent of the wage gap between white men and black women. In addition to the difference in age group, some of the difference between Corcoran's and Mincer and Polachek's findings with re- spect to the importance of the human capital variables may reflect a growing work force attachment of women over the period spanned by the 1967 NLS and the 1976 PSID (Mincer, 19791. These findings of large unexplained wage gaps for white and black women are impressive in light of Corcoran's detailed controls for productivity-related factors. In addition, since her findings are based on OLS estimates, she has not ad- justed for the simultaneous equations bias 24 Using longitudinal panel data from the NLS on the 30- to 44-year age group, Mincer and Ofek (1982) find direct evidence of rapid wage growth upon reentry. Unlike Corcoran (1979), however, they attribute it to the "repair" of previously eroded human capital. discussed in reference to the Mincer and Polachek (1974) study. This suggests that, at least with regard to this consideration, she has underestimated the effect of labor market discrimination. An additional issue raised by the human capital model is the interpretation of sex dif- ferences in returns to experience ant! mar- ital status. For example, Mincer and Pola- chek (1974, p. S103) argue that The association of lower [femaleJ coefficients with lesser work experience is not fortuitious: a smaller fraction of time and energy is devoted to job advancement (training, learning, getting ahead) per unit of time by persons whose work attach- ment is lower. Hence, the 45 percent figure in the explanation of the gap by duration-of-work experience alone may be viewed as an under- statement. To what extent do sex differences in returns to experience and marriage reflect employer discrimination and to what extent do they reflect women's choices? While a definitive answer is not available, the evidence sug- gests that discrimination most likely plays a role. With respect to returns to experience, a Stiffly by Duncan and Hoffman (1979) is par- ticularly interesting. Using direct measures of on-thejob training from the 1976 PSID, they find that men and women receive about the same payoff to on-thejob training. How- ever, "past years of work experience have a high payoff in training for men, especially white men, but have very little effect on the chances that women will receive training" (p. 6011. They see their results as consistent with a view that firms have different pro- motion practices on the basis of sex (and race). Thus, the observed lower returns to experience of women may reflect to some extent employer discrimination in permit- ting women access to on-the job training op- portunities. On the other hand, Sandell and Shapiro (1980), analyzing data from the NLS, have found that young white and black women who plan to work at age 35 have experience-
132 FRANCINE D. BLAU wage profiles that "begin at a lower point and have a steeper (initial) slope than those of their 'no-work-plans' counterparts" (p. 3431. Thus, some of the sex difference in work experience coefficients may reflect women's own investment decisions based on their work expectations. However, Sandall and Shapiro also find that the returns to job tenure (sen- iority), which they take as an indication of investment in firm-specific training, do not differ significantly between those who plan to work at age 35 and those who do not. This in turn suggests that part of the differential access to on-thejob training opportunities by sex implied by Duncan and Hoffman's findings may reflect the inability or the un- wiDingness of employers to distinguish be- tween the committed and uncommitted group. As our previous discussion suggests, there is some difference of opinion as to whether such statistical discrimination would represent eco- nomic discrimination or not. However, it is likely to result in committed women getting less training and receiving lower returns to job tenure than do committed men. In evaluating the implications of Sandell and Shapiro's findings for sex differences in earnings, it is instructive to consider Ar- row's (1973) notion of perceptual equilib- rium. Given the set of market opportunities she can reasonably expect and her expected value of nonmarket time, each young woman determines her work plans for age 35. Since women with given characteristics in the Sandell and Shapiro sample presumably face similar opportunities, the differences in their work plans probably reflect differences in their expected value of nonmarket time (perhaps due to differences in tastes). How- ever, this does not preclude the possibility that, if confronted with a different set of market opportunities, substantially more of the women would be committed to market work. It is even possible that, given the male set of job opportunities (with similar returns to experience and job tenure), they would exhibit the same degree of labor force com- mitment as that of males. With respect to worker and employer firm- specif1c training investment decisions, it is job (rather than labor force) stability that is the issue. Some evidence in favor of the Arrow notion in this case is provided by the findings of Blau and Kahn (1981), who used data on young men and young women from the NLS to analyze sex differences in quit- ting. They found that, all else equal (in- cluding job-related characteristics), white and black women were no more likely to quit their jobs than were men of the same race. They also found that a high proportion of the observed sex differential in quitting was associated with job rather than personal characteristics.25 Similar findings were re- ported by Viscusi (1980) for a larger age range from the PSID.26 Both Blau and Kahn and Viscusi found that, aD else equal, blacks were less likely to quit their jobs than white work- ers of the same sex. Blau and Kahn also report some support for one of the models of statistical discrimination presented ear- lier in that female quits were found to be less accurately predicted than men's by the explanatory variables.27 No basis for statis- tical discrimination was found in the case of race. With respect to the interpretation of sex 25 Job characteristics include current wage, long-run earnings opportunities associated with the job, collec- tive bargaining coverage, and industry and occupation dummy variables. 26 Ragan and Smith (1981) find that sex differences in industry turnover rates explain a substantial portion of the sex difference in earnings among individuals. However, as they acknowledge, since their data refer to the industry's history and not the individual's, their findings are consistent with the possibility that women are restricted to low wage/high turnover jobs (e.g., those requiring little specific training). Osterman (1979) found no sex differences in absenteeism, all else equal, for a sample of professional workers. 27 While Osterman (1979) reports no basis for statis- tical discrimination on this ground, Kahn (1981) shows that Osterman does not employ the correct indicator of predictability. When the correct indicator is used, Kahn finds that women's absenteeism is less accurately predicted than men's.
OCCUPATIONAL SEGREGATION AND LABOR MARKET DlSCRIMlNATION 133 differences in the returns to marital status, Hill (1979) uses explicit data on experience, human capital investment and labor force attachment from the 1976 PSID to explore this issue further. She finds that, all else equal, marriage has strong positive wage ef- fects for (white and black) men, while the effects of marriage on (white and black) women's wages are not significantly differ- ent from zero. Malkie! and Malkie! (1973) and Osterman (1979) report similar findings for the wage effects of marriage from studies of sex differences in the wages of profes- sional workers, also including good controls for work experience and attachment. A sup- ply-side explanation for these findings can- not be entirely ruled out: given the tradi- tional division of labor within the home, married men may be more highly motivated or hardworking than single men with similar measured characteristics, while married women may be less highly motivated or hardworking than their single counterparts. But, as Osterman points out, how do we then explain the absence of a negative effect of marriage on the earnings of women? Moreover, Hill examined the effect of mar- ital status and number of children on wages as more explicit controls for experience, hu- man capital investments, and labor force at- tachment were added to the wage regres- sions. She finds that the wage effect of marital status among all race/sex groups remains quite stable and concludes that marital status does not serve as a proxy for these productivity- related factors. If marital status is not serv- ing as a proxy for these obvious and impor- tant factors, it seems doubtful that it is serv- ing as a proxy for more subtle traits like motivation. However, Hill does find that number of children is to some extent proxy for these factors. Both Osterman and Hill feel that employers may believe that mar- ried men deserve higher salaries because of their greater financial responsibilities. Given traditional views of men's and women's eco- nomic roles, they do not feel the same is true of married women. Such a difference in treatment, if it exists, would have to be classed as discriminatory. The evidence reviewed here strongly sug- gests that labor market discrimination does indeed play a role in producing the observed male-female pay differential. While it is dif- ficult to pinpoint the exact portion of the sex pay gap due to discrimination, the findings suggest that over half of the clifferential can- not be explained by sex differences in pro- ductivity-related factors. Some considera- tions (e.g., omitted variables) suggest that this may be an overestimate of the magni- tude of discrimination. On the other hand, other considerations (e.g., the impact of la- bor market discrimination on the incentives to acquire experience, training, etc.) sug- gest that it may be an underestimate. Thus, we are still left with fairly strong evidence of the importance of discrimination. Occupational Segregation and Earnings Having determined the overall labor mar- ket context in which employment segrega- tion by sex takes place, we are now able to turn to a consideration of the empirical evi- dence regarding the role played by such seg- regation in producing pay differences and a consideration of some recent evidence on the causes of such segregation. According to the discrimination models considered earlier, segregation may work to Tower women's earn- ings because of a lesser provision of on-the- job training and fewer incentives for worker stability in female jobs (Arrow's perceptual equilibrium model, institutional models) and/ or because of overcrowding. While these no- tions suggest that it would be instructive to look at the relationship between earnings particularly the discriminatory pay gap and segregation, there are three major problems in arriving at an empirical estimate. The first is a data problem. Most data sources, even detailed census data, tend to aggregate some male and female jobs into apparently integrated categories (Hartmann and Reskin, 19821. Further, insofar as men
134 FRANCINE D. BLAU and women are segregated by firm within occupational categories, aggregation across firms will result in an underestimate of the extent to which men and women are in seg- regated work settings (Blau, 19771. The im- pact of both of these factors is revealed in a recent study by Bielby and Baron (this vol- ume) of 393 California establishments. Us- ing the employers' own job classifications, they found that 51 percent of the firms were completely sex-segregated with respect to job classification: no men and women shared the same job title. An additional 8 percent of the firms were single-sex establishments. The mean index of segregation of the re- maining 41 percent of firms was 84.1. It has been found that the portion of the sex pay gap associated with occupational differences is larger, the finer the occupational cate- gories employed (Treiman and Hartmann, 19811. Thus, it seems reasonable to con- clude that aggregation problems result in an underestimate of the impact of employment segregation on the sex pay gap. The second problem is more conceptual. The logical way to determine the impact of occupational characteristics or categories on pay, all else equal, is to control for sex or to look within sex groups. Yet it seems pos- sible that the existence of overcrowding in female jobs may lower the wages of women in male jobs. Even when women work in male jobs their opportunity set may differ from that of their male coworkers: the lower- paying alternatives they face in the female sector may reduce their supply price to the firm. This is a potential wage spflIover effect of overcrowding. Further, women may face discrimination in the male sector that lowers their relative wages there. Indeed, in Berg- mann's original formulation of the over- crowding model it is the exclusion of women from male jobs due to discrimination that causes the overcrowding of the female sec- tor. It makes intuitive sense that women remain concentrates! in female jobs because they have little to gain by obtaining male jobs. Thus, measuring the impact of occu- pation on pay by contrasting the wages of women in male ant! female jobs, celery par- ibus, may result in an underestimate. On the other hand, the internal labor market (institutional) mode} suggests that when women are able to obtain employment in male jobs within a firm, they should be paid at about the same rate as men. This implies that pay comparisons of women in male and female jobs can provide good estimates of the impact of occupation on wages. How- ever, since data are generally aggregated over firms, women may earn less than men do in male jobs if they work for low-paying firms. Further, women may be segregated by job level within the same occupational category. Judging the impact of occupation by com- paring the wages of men in male and female jobs does not appear to be satisfactory either. A problem here is that men are not believed to be discriminated against in or excluded from male jobs, so the question arises as to why they are employed in the female sector. If it is because they have very strong non- pecuniary (nonmonetary) tastes for female jobs, then their wages may reflect a differ- ential due to overcrowding. The pay differ- ential between men in male and female jobs may also reflect overcrowding if bad luck, poor information, or job rationing in male jobs (e.g., due to unions) are responsible for the employment of men in female jobs. On the other hand, men may work in female jobs because they have found a niche there that pays comparably with what they could earn in male jobs, e.g., due to employment in a high-paying firm or at a high level in the occupation hierarchy. In that case a comparison of men's wages in male and fe- male jobs wfl} not reflect an overcrowding differential. These two considerations suggest that empirical estimates may understate the con- tribution of employment segregation to the sex pay gap. A third point works in the op- posite direction. Workers in higher-paying jobs (or firms) may have unobserved char-
OCCUPATIONAL SEGREGATION AND LABOR MARKET DlSCRIMlNATION acteristics that are associated with higher productivity: what appear to be occupation effects on wages may actually be due to unobserved productivity differences among workers. The union-impact literature suggests some fruitful alternative approaches for examining the impact of crowding, since many analo- gous problems arise in investigating that is- sue. For example, papers by Kahn (1980) and Flanagan (1976) suggest that one might look at the effect on the wages of men and women in male and female jobs of changes over time or of differences across labor mar- kets, in the degree of crowding. This type of approach could provide an empirical es- timate of the impact of crowding that takes spillover effects into account. The selectivity problem of comparing the wages of women (or of men) in male and female jobs might be overcome by using a technique devel- opec} by Lee (1978) in his investigation of the union-nonunion differential. While existing studies may deal inade- quately with the problems raised here, it is still instructive to review the findings in this area. Using data from the 1980 census, Trei- man and Hartmann (1981) found that 35 to 39 percent of the earnings difference between men and women was associated with sex dif- ferences in the distribution of their employ- ment among 479 detailed categories.28 Oc- cupational differences appear to be a significant factor in explaining the sex pay gap, even when other productivity-related factors are con- trolled for. For example, Oaxaca (1973a,b) found that the inclusion of controls for major occupation and industry and for cuss of worker increased the portion of the sex pay gap ex- plained from 20 to 37 percent in the case of whites and from 6 to 39 percent in the case of blacks. In this case job characteristics ac- counted for some 20 to 35 percent of the dis- criminatory pay gap. Using 1950 census data, Sanborn (1964) was able to explain 43 percent 28 See also Chiswick et al. (1974~. 135 of the sex pay gap on the basis of controls for such factors as detailed occupation, age, and education.29 The greater magnitude of the ex- plained differential in the Sanborn study than in the studies of Fuchs (1971) and Oaxaca (1973a,b) cited earlier, in which occupational controls were not included, is an indication of the importance of occupational category in determining earnings. Further, using date from the 1974 PSID and the 1967 NLS, England (1981, 1982) found that, all else equal, the percentage of females in the occupation is sig- nificantly negatively related to female earn- ings.30 Table 7-1 illustrates the impact of occupa- tional category within an individual firm a large fiduciary institution.3i In equation (1), only controls for sex and race are entered into the regression, thus providing an estimate of the gross or unadjusted sex-race differentials. Equation (2) includes controls for productiv- ity-related individual characteristics but no controls for occupational characteristics. Ibis 29 Including adjustments for even more narrowly de- fined occupational categories from the BLS, as well as sex differences in turnover, absenteeism, and work ex- perience, Sanborn (1964) was able to explain 71 percent of the pay gap. The problem with this portion of his analysis is that he most probably engaged in double counting. He assumed that the sex differences in age and education that prevailed within the census cate- gories also prevailed within the more detailed BLS categories. Further, he adjusted within occupations for estimates of aggregate sex differences in turnover, ab- senteeism, and work experience. Leaving aside the issue of endogeneity, labor quality differences between men and women are likely to be considerably smaller within occupations than in the aggregate, since it is these traits that sort people into occupations. Indeed, in the presence of discrimination women may be more qualified than men in specific occupations. See Ham- ilton (1973) for some evidence consistent with the latter possibility. 30 See also Roos (1981), Ferber and Lowry (1976), Cabral et al. (1981), Stevenson (1975), and Jusenius (1977~. 3} This analysis was part of the statistical evidence developed by Janice Madden and me in an employ- ment discrimination case. Madden (1982) also provides a discussion and analysis of these data.
36 FRANCINE D. BLAU TABLE 7-1 Regression Analysis of the Salaries of Active Employees in 1978 (standard errors) Dependent Variable: Natural Log of Annual Salary Independent Variables (1) (2) (3) Personal characteristics Female Black Other minority Education (highest grade completed) Age (in years) Age (squared) Firm experience (in years) Firm experience (squared) Job characteristics Hay points Firm officer Other exempt Constant term R square F statistic Number of employees O. 5659a (0.0183) _ 0.3573a (0.0243) 0.2091a (0.0359) 9.7748 0.4156 476.0479a 2012 O. 3236a (0.0151) O. 1985a (0.0185) _ O. 0943a (0.0267) O. 09410a (0.0035) O. 0512a (0.0042) O. 00059a (0.00005) O. 0258a (0.0029) O. 00027a (0.00010) 7.1928 0.6873 550.1990a 2012 O. 0785a (0.0083) O. 0948a (0.0092) _ O. 0459a (0.0130) O. 0176a (0.0021) O. 0268a (0.0022) O. 00030a (0.00003) O. 0088a (0.0014) O. 00018a (0.00005) O. 0007a ~o.ooool) O. 3628a (0.0132) O. 2332a (0.0116) 8.2253 0.9294 1806 a Significant at the 1 percent level on a two-tailed test. gives the total effect of discrimination (oper- ating both through unequal pay for equal work and unequal access to higher-paying jobs). Discrimination is estimated to account for 57 percent of the gross pay differential in the case of women (controlling for race) and 45 to 56 percent of the gross pay differential in the case of blacks and other minorities (controlling for sex). Controlling for occupational character- istics in equation (3) gives us an estimate of pay differences between equally qualified male and female (black and white) workers in sim- ilar job categories. In this case the job char- acteristics include hay points an employer evaluation of the value ofthe job to the firm and two dummy variables indicating whether the individual is an officer of the firm or is in another managerial or professional occupation (exempt from the Fair Labor Standards Act). A relatively small proportion of the discrim- inatory sex differential (controlling for race)- 24 percent (-0.0785/-0.3236) is due to pay differences within similar occupational categories. The remainder, 76 percent, is due to sex differences in distribution among oc- cupational categories within the firm. Occu- pational differences explain somewhat less than half of the discriminatory pay gap in the case of blacks and other minorities (controlling for sex). There are some problems with these spec- ifications. For one thing age rather than actual labor market experience is used as an explan- atory variable due to data availability. (The
OCCUPATIONAL SEGREGATION AND LABOR MARKET DISCRIMINATION 137 firm experience variable does, however, measure actual firm experience.) However, the studies reviewed earlier strongly suggest that the discriminatory differential would per- sist even if we were able to control for actual labor market experience. For another, job grade Way points) rather than occupational category is employed, making this similar to the type of investigation one would undertake in a study of the issue of comparable worth. But there are some advantages to the use of hay points as an overall measure of job level that cuts across male and female jobs. It over- comes one of the practical problems with ef- forts to ascertain the size of intraoccupational sex pay differentials: paucity of data on one sex group or another within a job category due to the very sex segregation by occupation that we seek to study.32 These results support the notion that studies of the impact of oc- cupation undertaken at the level of the firm and utilizing job categories more closely ap- proximating the job titles used by the em- ployer will reveal a greater impact of job cat- egory on wages than aggregate analyses. The Causes of Occupational Segregation As discussed earlier, the human capital model provides an explanation for occupa- tional segregation by sex in terms of wom- en's optimizing behavior, given the tradi- tional division of labor by sex within the fam fly. Polachek (1979, 1981) provides some support for this view when jobs are cate- gorized according to variants of the census major occupational groups. The problem with his approach is that these major occupa- tional categories combine predominantly male and predominantly female jobs (Eng- land, 19821. England (1981, 1982) explicitly examines whether women's earnings pat- terns in predominantly female and predom- 32 Employer job evaluation schemes may, however, understate the relative value of predominantly female occupations (Treiman and Hartmann, 1981~. inaptly male occupations differ in the way predicted by the human capital model. She finds that the earnings of women in pre- dominantly female occupations do not show lower rates of either depreciation or appre- ciation than do the earnings of women in occupations employing more males.33 Fur- ther, she finds women who have discontin- uous work histories are no more likely to be in predominantly female occupations than are women who have been employed more continuously. Indeed, since she finds that women earn less in female jobs at all levels of experience, she concludes that "the evi- dence does not support the contention of human capital theorists that women maxi- mize lifetime earnings by choosing female occupations" (England, 1981, p. 18~. If we provisionally conclude that the hu- man capital analysis of occupational choice discussed above does not explain occupa- tional segregation, at least at the aggregate level, what does? Many potential candidates remain, ranging from premarket discrimi- nation (e.g., by families through the social- ization process, or by schools through the actions of teachers, guidance counselors, or admissions committees) to the exclusionary practices of employers (due to their own tastes, statistical discriminations, and/or the tastes of employees or customers). Clearly, considerable additional work needs to be done to narrow the field and/or to attach relative weights to these competing expla- nations. Pay Differentials Within Occupations The relatively flat earnings profiles of women in female jobs are consistent with 33 See also Beller (1982~. She finds that when one examines detailed (three-digit) census occupations, the evidence for the human capital model is mixed in that the expected signs on the labor supply variables are not always obtained. King (1977) finds little evidence of flatter age earnings profiles for women in female as compared with male professions.
138 FRANCINE D. BLAU the notion advanced in the institutional view and implied by Arrow's notion of perceptual equilibrium that employers would structure female jobs to fit the average perceived characteristics of women workers. How- ever, the finding that women in male jobs also have relatively flat earnings profiles might at first appear inconsistent with the notion that the internal labor market mandates equal pay for equal work by sex. However, it should be noted that the census categories are ag- gregated across job levels and firms. Thus, firms may pay women and men in the same job category at roughly the same rate, but promote women more sIowly.34 Further, it may be that women and men in the same census job category are segregated by firm. Blau (1977) examines the extent of employ- ment segregation by sex within occupational categories and its relationship to intraoc- cupational pay differentials within the con- text of an institutional model. Her findings suggest that pay differentials between men and women in the same occupational cate- gory may reflect hiring discrimination by firms. marily sought by and attracted to the higher- wage establishments, while female workers for the most part find employment in the lower-paying firms, which, regardless of their preferences, are less able to compete for male labor. Blau tests this mode} using unpublished 1970 wage data from the Bureau of Labor Statistics on extremely narrow, white-collar occupational categories (e.g., accounting clerk, class A) in three northeastern cities. She argues that within such narrow cate- gories male and female labor is likely to be fairly homogeneous.35 Blau finds that within occupations men and women are segregated by establishments to an extent in excess of what would be expected on the basis of chance. Within firms, occupational pay dif- ferences are found to be relatively small, and sex pay differentials within occupations are primarily due to differences in pay rates among (rather than within) firms.36 Further, men tend to earn less when they work with women, which is counter to what we would expect on the basis of the Becker mode! if discriminatory tastes were located in em- Blau postulates that institutional and mar- ployees. - ket forces determine a wage hierarchy of Blau finds evidence of a wage hierarchy firms within the local labor market that is consistent across occupational categories. She argues that, while employer tastes for dis- crimination against women are fairly wide- spread, the ability to exercise them is con- strained by the firm's position in the wage hierarchy. That position is determined by a variety of factors and cannot easily be al- tered to accommodate employer prefer- ences regarding the sex composition of spe- cific occupational categories. ~us, in each occupational category male workers are pri- of firms that is consistent across occupations and sex groups. Controlling for occupational mix, the representation of women in the firm is found to be consistent across occu- pations and inversely related to the wage standing of the firm. Note that these find- ings also conflict with the Becker model. In the case of employee preferences it is not expected that men will earn more when they work with relatively fewer women. In the case of employer preferences it is not ex- pected that the firms that hire relatively the 34 Note that such a sex difference by job level would not support the human capital view in that it would not be economically rational for women to opt to take the lower-paid training positions, but not to reap the gains of moving up the job ladder. For findings sug- gesting that women have lower promotion probabili- ties, see Duncan and Hoffman (1979), Cabral et al. (1981), and Malkiel and Malkiel (1973~. 35 See footnote 29. 36 For other studies reporting differences in the dis- tribution of men and women by firm that are associated with pay differentials, see Buckley (1971), McNulty (1967), Bridges and Berk (1974), Talbert and Bose (1977), Allison (1976), and Dussault and Bose-Lizee (1980~.
OCCUPATIONAL SEGREGATION _ LABOR MARKET DISCRIMINATION 139 fewest women (presumably the most dis- criminatory firms) will pay women the high- est wage rates.37 CONCLUSIONS Various explanations have been offered for the pay and occupational differences between male and female workers. Some emphasize labor market discrimination, while others, most notably the human capital model, focus on the voluntary choices of women. A review of the empirical literature strongly suggests that, all else equal (including fairly refined meas- ures of work experience and labor force at- tachment), women do earn less Han men. This suggests that labor market discrimination does indeed play a role in producing the lower earnings of women. However, we lack a widely accepted economic theory of the role of oc- cupational segregation in producing this dif- ferential and of the persistence of sex discrim- ination in the labor market over time in the face of competitive forces. Perhaps it is time now to devote less of our empirical efforts to ascertaining the existence of discrimination and more toward determining which mode! of dis- crimination is most consistent with the data and the mechanisms by which these discrim- inatory outcomes are produced. On the basis of the existing evidence it appears that sex segregation in employment is an important mechanism for producing sex differences in earnings and that the occupational differences between men and women do not seem to be consistent with optimizing behavior on the part of women. However, considerably more work is needed to understand the causes of sex differences in occupational distributions fully and to determine the role of such oc- 37 Note that the comparisons made in the text be- tween the results expected on the basis of the Becker and institutional models rest fairly heavily on the as- sumption that labor is fairly homogeneous within these detailed occupations. Otherwise, variations in labor quality might account for these interfirm differences in 1973 pay rates. cupational differences in producing male-fe- male pay clifferentials. One area of particular concern is the issue of the impact of crowding in female jobs on the wages of women in male jobs. Finally, the question of the indirect ef- fects of discrimination on the qualifications of women (and minorities) is another area upon which future research could fruitfully be fo- cused. ACKNOWLEDGMENTS Portions of this chapter draw on my work entitled "Discrimination Against Women: Theory and Evidence," in William A. Dar- ity, Jr., e(l., Labor Economics: Modern Views (Boston: Kluwer-Nijhoff, 1984~. I am grate- ful to the publisher for granting permission for the reproduction of some of that material here. I am grateful for the comments and sug- gestions of Barbara Reskin, William Darity, Ir., Katherine Abraham, Andrea Beller, Barbara Bergmann, Charles Brown, Paula England, Marianne Ferber, Claudia Gol- din, Joan Huber, Robert Hutchens, Law- rence Kahn, Michael Reich, Harvey Rosen, and two anonymous referees. I am indebted to Susan Schwochan for excellent research assistance. REFERENCES Aigner, D. J., and G. C. Cain 1977 "Statistical Theories of Discrimination in La- bor Markets," Industrial and Labor Relations Review, 30, 2 Jan.) 175-187. Allison, E. K. 1976 "Sex Linked Earnings Differentials in the Beauty Industry," Journal of Human Resources, 11, 3 (Summer) 383-390. Arrow, K. 1972a "Models of Job Discrimination," in A. H. Pas- cal, Ed., Racial Discrimination in Economic Life (Lexington, Mass: D. C. Heath and Co.) 83-102. 1972b "Some Mathematical Models of Race in the Labor Market," in A. H. Pascal, Ed., Racial Discrimination in Economic Life (Lexington, Mass: D. C. Heath and Co.) 187-204. "The Theory of Discrimination," in O. Ashen- felter and A. Bees, Eds., Discrimination in La-
40 bar Markets (Princeton, New Jersey: Princeton University Press) 3~3. 1976 "Economic Dimensions of Occupational Seg- regaffon: Comment I," Signs, 1, 3, Part 2 (Spring), 233-237. Baker, C. O. Undated Earned Degrees Conferred: An Examination of Recent Trends, National Center for Edu- caffon Staffsffcs. Barnes, W. F., and E. Jones 1974 "Differences in Male and Female Quitting," Journal of Human Resources, 9, 4 (Fall), 439- 451. Becker, G. 1957 The Economics of Discrimination (Chicago: University of Chicago Press). 1965 "A Theory of the Allocation of Time," Eco- nomic Journal, 75, 229 (Sept.), 493-517. 1973 "A Theory of Marriage," Journal of Political Economy, 81, 4 Quly/Aug.), 813-846. Belter, A. H. 1982 "Occupational Segregaffon by Sex: Determi- nants and Changes," Journal of Human Re- sources, 17, 3 (Summer), 371-392. Bergmann, B. R. 1974 ''occupational Segregaffon, Wages and Profits When Employers Discriminate by Race or Sex," Eastern Economic Journal, 1 (Apr.-July) 103- 110. 1976 "Reducing the Pervasiveness of Discrimina- ffon,"inE. Ginzberg, Ed., Jobsfor Americans (Englewood Cliffs, N.J.: Prentice Hall) 120- 141. Bergmann, B. R., and W. Darity, Jr. 1981 "Social Relations in the Workplace and Em- ployer Discrimination," Proceedings of the Thirty-Third Annual Meeting of the Industrial Cain, ' FRANCINE D. BLAU in the Labor Market: An Appraisal," Signs: Journal of Women in Culture and Society, 1, 3, Part 2 (Spring), 181-199. Blau, F. D., and L. M. Kahn 1981 "Race and Sex Differences in Quits by Young Workers," Industrial and Labor Relations Re- view, 34, 4 Quly) 563-577. Blinder, A. S. 1973 "Wage Discrimination: Reduced Form and Structural Estimates," Journal of Human Re- sources, 8, 4 (Fall) 436~55. Bridges, W. P., and R. A. Berk 1974 "Determinants of White-Collar Income: An Evaluation of Equal Pay for Equal Work," So- cial Science Research, 3, 3 (Sept.), 211-234. Brown, C. 1981 "Black/White Earnings Ratios Since the Civil Rights Act of 1964: The Importance of Labor Market Drop-outs," NBER Working Paper #617 Oan.~. 1982 "The Federal Attack on Labor Market Dis- crimination: The Mouse That Roared?" Re- search in Labor Economics, 5, 33-68. Buckley, J. E. 1971 "Pay Differences Between Men and Women in the Same Job," Monthly Labor Review, 94, 11 (Nov.), 36-39. Butler, R. J. 1982 "A Note on Estimating Wage Discrimination in the Labor Market," Journal of Human Re- sources, 17, 4 (Fall), 606 621. Cabral, R., M. A. Ferber, and C. A. Green 1981 "Men and Women in Fiduciary Institutions: A Study of Sex Differences in Career Develop- ment," Review of Economics and Statistics, 63, 4 (Nov.) 573-580. Relations Research Association, University of 1976 "The Challenge of Segmented Labor Market Wisconsin, Madison, 155-162. Blau, F. D. 1975 "Longitudinal Patterns of Female Labor Force Participation," in H. S. Parnes et al., Dual Careers: A Longitudinal Analysis of the Labor Market Experience of Women, V. 4 (Colum- bus, Ohio: Center for Human Resource Re- search, Ohio State University, Dec.), 27-55. 1977 Equal Pay in the Office (Lexington, Mass: D. C. Heath and Co.~. 1978 "The Impact of the Unemployment Rate on Labor Force Entries and Exits," in Women's Changing Roles at Home and on the Job. A special report of the National Commission for Manpower Policy, Report No. 26 (Sept.), 263- 286. Blau, F. D., and W. E. Hendricks 1979 "Occupational Segregation by Sex: Trends and Prospects," Journal of Human Resources, 14, 2 (Spring), 197-210. Blau, F. D., and C. L. Jusenius 1976 "Economists' Approaches to Sex Segregation Theories to Orthodox Theory: A Survey," Jour- nal of Economic Literature, 14, 4 (Dec.), 1215- 1257. Cardwell, L. A., and M. R. Rosenzweig 1980 "Economic Mobility, Monopsonistic Discrim- ination and Sex Differences in Wages," South- ern Economic Journal, 4~;, 4, (Apr.), 1102-1117. Chiplin, B. 1981 "An Alternative Approach to the Measurement of Sex Discrimination: An Illustration From University Enhance," Economic Journal, 91, 394 (Dec.), 988-997. Chiswick, B. R., J. A. O'Neill, J. Faclder, and S. W. Polachek 1974 "The Effect of Occupation on Race and Sex Differences in Hourly Earnings," Business and Economic Statistics Section Proceedings of the American Statistical Society, 219-228. Cogan, J. 1980 "Married Women's Labor Supply: A Compar- ison of Alternative Estimation Procedures," in J. P. Smith, Ed., Female Labor Supply: The-
OCCUPATIONAL SEGREGATION AND LABOR MARKET DlSCRIMlNATlON 141 ory and Estimation (Princeton, N.J.: Princeton Freeman, R. B. University Press) 90-118. 1973 "Changes in the Labor Market for Black Amer- icans, 1948-1972," Brookings Papers On Eco- nomic Activity, 67-131. Fuchs, V. R. 1971 "Differences in Hourly Earnings Between Men and Women," Monthly Labor Review, 94, 5 (May), 9-15. Gordon, D. M. 1972 Theories of Poverty and Underemployment (Lexington, Mass: D. C. Heath and Co.~. Gronau, R. 1973 "The Intrafamily Allocation of Time: The Value of Housewives' Time," American Economic Review, 63, 4 (Sept.), 634-651. "Wage Comparisons A Selectivity Bias," Journal of Political Economy 82, 6 (Nov./Dec.), 1119-1143. Gwartney, J. D., and R. Stroup 1973 "Measurement of Employment Discrimination According to Sex," Southern Economic Jour- nal, 39, 4 (Apr.), 575-587. Hamilton, M. T. 1973 "Sex and Income Inequality Among the Em- ployed," Annals of the American Academy of Political and Social Sciences, 409 (Sept.), 42- 52. Hartmann, H. I., and B. Reskin 1982 "Job Segregation: Trends and Prospects," pa- per presented at a Conference on Occupaffona Segregation, Ford Foundation, New York City, June. Hashimoto, M., and L. Kochin 1980 "A Bias in the Statistical Estimation of the Ef- fects of Discrimination," Economic Inquiry, 18 duly), 478486. Corcoran, M. 1978 "The Structure of Female Wages," The Amer- ican Economic Review, papers and proceed- ings, 68, 2 (May), 165-170. 1979 "Work Experience, Labor Force Withdrawals and Women's Earnings: Empirical Results Us- ing the 1976 Panel Study of Income Dynam- ics," in C. B. Lloyd, E. Andrews, and C. L. Gilroy, Eds., Women in the Labor Market (New York: Columbia University Press). Darity, W. A., Jr. 1982 "The Human Capital Approach to Black-White Earnings Inequality: Some Unsettled Ques- tions," Journal of Human Resources, 17, 1 1974 (Winter), 72-93. Darity, W. A., Jr., and S. Myers, 1980 "Changes in Black-White Income Inequality, 1968-1978: A Decade of Progress," Review of Black Political Economy, (Summer), 354-379. Doeringer, P. B., and M. J. Piore 1971 internal Labor Markets and Manpower Anal- ysis (Lexington, Mass: D. C. Heath and Co.~. Duncan, G. J., and S. Hoffman 1979 "On-the-Job Training and Earnings Differ- ences by Race and Sex," Review of Economics and Statistics, 61, 4 (Nov.), 594-603. Dussault, G., and R. Rose-Lizee 1980 "La Discrimination a l'Egard des femmes et la Segregation sur les Marches du Travail: Le Cas des Employes de Bureau a Montreal," in La Discrimination a l'Egard des femmes sur le Marche du Travail, Collection Tires a Part, No. 25 (Quebec: Universite Laval) 475-504. Economic Report of the President 1974 (Wash~ngton, D.C.: U.S. Govt. Printing Of- Heckman, J. J. Edgeworth, F. Y. 1922 "Equal Pay to Men and Women for Equal Work," Economic Journal, 32 (Sept.) 431457. England, P. 1981 "Wage Appreciation and Depreciation: A Test of Neoclassical Economic Explanations of Oc- cupational Sex Segregation," mimeo (Dec.~. 1982 "The Failure of Human Capital Theory to Ex- plain Occupational Sex Segregation," Journal of Human Resources, 17, 3 (Summer), 358- 370. Fawcett, M. G. 1918 "Equal Pay for Equal Work," Economic Jour- nal, 28 (March), 1-6. Ferber, M. A., and H. M. Lowry 1981 1976 "The Sex Differential in Earnings: A Reap- praisal," Industrial and Labor Relations Re- view, 29, 3 (Apr.), 377-387. Flanagan, R. J. 1976 "Wage Interdependence in Unionized Labor Markets," Brookings Papers on Econom~c Ac- tivity, No. 3. 1974 "Shadow Prices, Market Wages and Labor Supply," Econometrica, 42, 4 July), 679-694. Hill, M. S. 1979 "The Wage Effects of Marital Status and Chil- dren," Journ~il of Human Resources, 14, 4 (Fall), 579-594. 1977 "The Influence of Work Experience, Skill Re- quirement, and Occupational Segregation on Women's Earnings," Jourruil of Econom~cs and Business, 29, 2 (Winter), 107-115. Kahn, L. M. 1980 "Union Spillover Effects on Unorganized La- bor Markets," Journal of Human Resources, 15, 1 (Winter), 87-98 "Sex Discrimination in Professional Employ- ment: A Case StudyComment," Industrial and Labor Relations Revi~v, 34, 2 Qan.), 273- 275. Kamalich, R. F., and S. W. Polachek 1982 "Discrimination: Fact or Fiction? An Exami- nation Using an Alternative Approach," South- ern Economic Journal, 49, 2 (Oct.).
42 King, A. G. 1977 "Is Occupational Segregation the Cause of the Flatter-Experience-Earnings Profiles of Women?" Journal of Human Resources, 12, 4 (Fall), 541-549. Kohen, A; I., with S. C. Breinich, P. Shields 1975 "Women and the Economy: A Bibliography and a Review of the Literature on Sex Differ- entiation in the Labor Market," Center for Hu- man Resource Research, College of Adminis- trative Science, Ohio State University. Insides, E. M. 1977 "Sex Differences in Wages and Employment: A Test of the Specific Capital Hypothesis," Economic Inquiry, 15, 4 (Oct.), 523-538. Lee, L. 1978 "Unionism and Wage Rates: A Simultaneous Equations Model With Qualitative and Lim- ited Dependent Variables," International Eco- nomic Review, 19 Qune), 415-433. Lewin P., and P. England 1973b 1982 "Reconceptualizing Statistical Discrimination," paper presented at the annual meetings of the Southwest Social Science Association, San An- tonio, Tex., March. Lloyd, C. B., and B. T. Niemi 1979 The Economics of Sex Differentials, (New York: Columbia University Press). Madden, J. F. 1973 The Economics of Sex Discrimination (Lexing- ton, Mass: D. C. Heath and Co.~. 1976 "Economic Dimensions of Occupational Seg- regation: Comment III," Signs, 1, 3, Part 2 (Spring), 245-250. 1982 "The Measurement of Employment Discrim- ination: Reduced Forms, Reverse Regression, Comparable Worth and the Definition of La- bor Markets," Proceedings of the Social Sta- tzstics Section at the Annual Meeting of the American Statistical Association, Cincinnati, Ohio, Aug. Malkiel, B. G., and J. A. Malkiel 1976 1973 "Male-Female Pay Differentials in Professional Employment," American Economic Review, 63, 4 (Sept.), 693-705. McNulty, D. J. 1967 "Differences in Pay Between Men and Women Workers," Monthly Labor Review, 90 (Dec.), 40-43. Medoff, J. L., and K. G. Abraham 1980 "Experience, Performance, and Earnings,~ Quarterly Journal of Economics, 95, (Dec. ), 703-736. 1981 "Are Those Paid More Really More Produc- tive? The Case of Experience," Journal of Hu- man Resources, 16, 2 (Spring) 186-216. Mincer, J. 1962 "Labor Force Participation of Married Women," in NBER, Aspects of Labor Economics (Prince- ton: Princeton University Press) 63-97. FRANC1NE D. BLAU ,. 1981 1979 "Wage Differentials: Comment," in C. B. Lloyd, E. S. Andrews, and C. L. Gilroy, Eds., Women in the Labor Market (New York: Columbia University Press) 278-285. Mincer, J., and H. Ofek 1982 "Interrupted Work Careers: Depreciation and Restoration of Human Capital," Journal of Hu- man Resources, 17, 1 (Winter), 3-24. Mincer, J., and S. Polachek 1974 "Family Investments in Human Capital: Earn- ings of Women," Journal of Political Economy, 82, 2 (Part 2) (March/Apr.), S76-S108. 1978 "An Exchange: Theory of Human Capital and the Earnings of Women: Women's Earnings Reexamined," Journal of Human Resources, 13, 1 (Winter), 118-134. Oaxaca, R. 1973a "Male-Female Wage Differentials in Urban Labor Markets," International Economic Re- view, 14, 3 (Oct.), 693-709. "Sex Discrimination in Wages," in O. Ashen- felter and A. Bees, Eds., Discrimination in Labor Markets (Princeton, N. J.: Princeton University Press) 124-151. Osterman, P. 1979 "Sex Discrimination in Professional Employ- ment: A Case Study," Industrial and Labor Relations Review, 32, 4 Duly), 451-464. Phelps, E. S. 1972 "The Statistical Theory of Racism and Sexism, American Economic Review, 62, 4 (Sept), 659- 661. Piore, M. I. 1971 "The Dual Labor Market: Theory and Impli- cations," in D. M. Gordon, Ed., Problems in Political Economy: An Urban Perspective (Lex- ington, Mass: D. C. Heath and Co.) 90-94. Polachek, S. W. 1975 "Potential Biases in Measuring Male-Female Discrimination,"Journal of Human Resources, 10, 2 (Spring), 205-229. "Occupational Segregation: An Alternative Hypothesis," Journal of Contemporary Busi- ness, 5, 1 (Winter), 1-12. 1979 "Occupational Segregation Among Women: Theory, Evidence, and A Prognosis," in C. B. Lloyd, E. S. and C. L. Gilroy, Eds., Women in the Labor Market (New York: Columbia University Press) 137-157. "Occupational Self-Selection: A Human Cap- ital Approach to Sex Differences in Occupa- tional Structure," Review of Economics and Statistics, 63, 1 (Feb.), 60-69. Ragan, J. F., Jr., and S. P. Smith 1981 "The Impact of Differences in Turnover Rates on Male/Female Pay Differentials," Journal of Human Resources, 16, 3 (Summer), 343-365. Roberts, H. V. 1980 "Statistical Biases in the Measurement of Em- ployment Discrimination," in E. R. Livernash,
OCCUPATIONAL SEGREGATION AND LABOR MARKET DlSCR]MlNATION 143 Ed., Comparable Worth: Issues and Alterna- tives (Washington, D.C.: Equal Employment Advisory Council) 173-196. Robinson, J. 1933 The Economics of Imperfect Competition (Lon- don: Macmillan and Co., Ltd.~. Boos, P. A. 1981 "Sexual Stratification in the Workplace: Male- Female Differences in Economic Returns to Occupation," Social Science Research, 10, 3 (Sept.), 195-224. Sanborn, H. 1964 "Pay Differences Between Men and Women," Industrial and Labor Relations Review, 17, 4 duly), 534-550. Sandell, S. H., and D. Shapiro 1978 "An Exchange: Theory of Human Capital and the Earnings of Women: A Reexamination of the Evidence," Journal of Human Resources, 13, 1 (Winter), 103-117. 1980 "Work Expectations, Human Capital Accu- mulaffon, and the Wages of Young Women," Journal of Human Resources, 15, 3 (Summer), 335-353. Sawhill, I. V. 1973 "The Economics of Discrimination Against Women: Some New Findings,"Journal of Hu- man Resources, 8, 3 (Summer), 383-396. Spence, A. M. 1973 "Job Market Signaling," Quarterly Journal of Economics, 87, 3 (Aug.), 355-374. 1974 Market Signaling (Cambridge, Mass: Harvard University Press). Stevenson, M. H. 1975 "Relative Wages and Sex Segregation by Oc- cupaffon," C. Lloyd, Ed., Sex, Discrimination, and the Division of Labor (New York: Colum- bia University Press). Strober, M. H. 1976 "Toward Dimorphics: A Summary Statement to the Conference on Occupational Segrega- ffon," Signs, 1, 3, Part 2 (Spring), 293-302. Suter, L., and H. Miller 1973 "Income Differences Between Men and Ca- reer Women," American Journal of Sociology, 78, 4 Jan.), 962-974. Talbert, J., and C. E. Bose 1977 "Wage Attainment Processes: The Retail Clerk Case," American Journal of Sociology, 83, 2 (Sept.), 403-424. Iturow, L. C. 1975 Generating Inequality (New York: Basic Books, Inc.~. Treiman, D. J., and H. I. Hartmann, Eds. 1981 Women, Work, and Wages: Equal Pay for Jobs of Equal Value (Washington, D.C.: National Academy Press). U.S. Department of Commerce, Bureau of the Census 1980 A Statistical Portrait of Women in the United States: 1978, Current Population Reports, Special Studies, Series P-23, No. 100 (Feb.~. 1982 Money Income and Poverty Status of Families and Persons in the United States: 1981 (Ad- vance Report), Current Population Reports Se- ries P-60, No. 134 (Washington, D.C.: U.S. Government Printing Office, July). U.S. Department of Labor, Bureau of Labor Statistics 1980 "Perspectives on Working Women," June. 1981 News Release 81-522 (Nov. 15~. 1982 Employment Earnings (Apr.~. U. S. Department of Labor, Employment and Training Administration 1981 Employment and Training Report of the Pres- ident. Viscusi, W. K. 1980 "Sex Differences in Worker Quitting," Review of Economics and Statistics, 62, 3 (Aug.) 388- 398. Wachter, M. 1974 "Primary and Secondary Labor Markets: A Cri- tique of the Dual Approach," Brookings Papers on Economic Activity, no. 3, 637-694. Weiss, Y., and R. Gronau 1981 "Expected Interruptions in Labour Force Par- ticipation and Sex-Related Differences in Earnings Growth," Review of Economic Stud- ies, 58 (Oct.), 607-619. Zellner, H. 1975 "Abe Determinants of Occupational Segrega- tion," in C. B. Lloyd, ea., Sex, Discrimination and the Division of Labor (New York: Colum- bia University Press) 125-145.