Revue économique | 245-256
Distribution électronique Cairn pour les éditions Presses de Sc. Po.. © Presses de Sc. Po.. Tous droits réservés pour tous pays. Il est interdit, sauf accord préalable et écrit de l’éditeur, de reproduire (notamment par photocopie) partiellement ou totalement le présent article, de le stocker dans une banque de données ou de le communiquer au public sous quelque forme et de quelque manière que ce soit.
Female Progress and Discrimination
Sandra E. Black
In the last 20 years, women's economic progress has been staggering. In 1979, the female/male ratio of median weekly earnings of full-time workers in the United States was about .6; by 1998, this number had grown to almost .74. These same numbers for France were .8 in 1979 to .9 in 1998. (Blau and Kahn ). This progress is particularly startling given that this has also been a period of marked increases in income inequality overall, declining relative wages of blacks, and declining real wages of low-skilled workers.
The question then becomes: What can explain this improvement in women's economic status? And how have changing environments affected the ability of firms to discriminate against women? Recent work has proposed a number of possible explanations for the progress of women, including changing social norms (in part facilitated by technological/pharmaceutical advances), increasing skill acquisition, changes in the general wage structure, and changes in the discriminatory environment. This paper examines the literature in this area, focusing first on the trends in women's economic status and how we can explain them and then turning to how the literature on discrimination has evolved and how it can inform our discussion.
It seems clear from the data that women have made significant progress over the past three decades. Figure 1 tracks the rapid increase in the median female/ male wage in the United States since 1979. At the same time, there was a drastic improvement in women's occupational status. Figure 2 shows the fraction of women in professional occupations relative to the population. Of particular interest is the fact that this trend is not driven by an increase in the percentage of women in “traditionally female” professions such as nursing and teaching; as the figure shows, the trend is quite similar even when one excludes teachers and nurses from the definition of “professional”.
As a final piece of evidence of women's improving economic status, there has been a striking trend in higher education enrollment; While figure 3 makes it clear that there has been an increasing fraction of women in college since the early 1960s, it also shows that, as of 1980, women have actually overtaken men in terms of representation in post-secondary institutions.
Some argue that one impetus for these trends was the introduction and spread of birth control to all women. Recent work by Goldin and Katz  has examined the role of contraception in women's labor market progress, arguing that the diffusion of birth control pills (“the pill”) enabled women to delay marriage and thereby invest more in their career. Despite the fact that the pill was approved in 1960 by the Food and Drug Administration and diffused rapidly among married women, access for single women was limited until years later when state law changes reduced the age of majority and extended mature minor decisions. Because of this variation in legislation across states over time, the authors are able to identify the effect of the pill on the age at first marriage among college women. The ease and efficacy of birth control pills reduced the likelihood that women engaging in sexual activity would become pregnant and thereby drop out of the labor market and get married. As a result, expected returns on investments in education were higher and more women decided to invest. Their evidence also shows the sharp increase in enrollment in graduate professional schools among women at this time, supporting their hypothesis. In addition, Goldin and Katz argue that the pill created a social multiplier effect by facilitating the general delay of marriage and thus increasing a woman's likelihood of finding an appropriate mate after school.
If the proliferation of birth control pills improved the returns to investment in human capital, then we should see an improvement in women's skills during this time period. Blau and Kahn  decompose the trends in women's relative wages in the United States and find just that. In their work, they attempt to disentangle the change in relative wages that was induced by changes in the overall wage structure (and would have occurred in the absence of any changing human capital accumulation of women) and the change that can be attributed to gender-specific factors such as differences in skills between men and women. Changes in the overall wage structure can relatively help or hurt women depending on where they are in the distribution. For example, women are disproportionately less educated and have fewer years of experience than men; during this time period, we witnessed an increase in the returns to education and experience which would disproportionately hurt women, had their education and experience remained the same. The authors conclude that rising inequality did in fact slow women's progress during the 1980s, reclaiming about 1/3 of the potential relative wage gains. On the other hand, this loss was offset by improvements in gender-specific factors. During this time period, women experienced increases in their occupational status, increases in relative experience levels, and a decline in the unexplained portion of the pay gap, which may be due in part to reduced discrimination.
Taking a slightly different perspective, Devereux  directly examines how industry growth and decline during the 1980s affected the gender wage gap in an effort to isolate the specific effect of changes in industrial composition on women. He exploits the panel data nature of the Panel Study of Income Dynamics to examine how individual-level wage changes over the decade are related to changes in employment in the industry the individual was in at the beginning of the decade. He also looks at the effects of employment changes in industries that the individual is predicted to be likely to move to based on their personal characteristics. He finds that women began the decade in industries that would subsequently grow much faster than industries that men were in initially. Also, individuals initially in industries that grow over the decade have much faster wage growth over the decade than individuals initially in slower-growing industries. Putting these two findings together, he concludes that changes in industry composition can account for most of the within-cohort increase in the wages of women relative to men during the 1980s.
Work by Black and Juhn  tries to understand what motivated the rise in the skills of women, focusing on the increasing fraction of women professionals. They discuss the possibility that women have responded to rising skill premiums by increasing their labor force participation. In 1970, less than 60 percent of college-educated women were working. The subsequent economy-wide increase in skill demand may have attracted educated women not only from other occupations, but from non-participation as well. Since almost all college educated men work, labor-market participation is less likely to be a factor for men. Figure 4 provides some evidence that this may be true.
Finally, Blau and Kahn  examine women's economic progress from an international perspective. They suggest that women in other countries have experienced very similar trends in human capital accumulation and economic progress. The primary disparity in the experiences between women in the United States and other countries is the large increase in income inequality in the United States in the 1980s that has hampered the progress of U.S. women. Because other countries, particularly those in Europe, did not experience the same rising inequality (primarily due to differing institutional settings), women in those countries fared better.
Even in the face of great improvements in the economic status of women, one still worries about the presence of gender discrimination in the market. Blau and Kahn  note that, even though the unexplained gender gap in 1988 is considerably smaller when one controls for all observable characteristics, including occupation, industry, and union status, a substantial portion of the pay gap “remained unexplained and potentially due to discrimination.” (page 82). But how have changing economic conditions during this time period affected the ability of firms to discriminate? Evidence suggests that there has been a reduction in the residual gender wage gap, which may suggest a decline in discrimination. Unfortunately, discrimination is very difficult to measure. This section of the paper will examine the recent literature on discrimination, focusing primarily on the effects of increasing competition, both as a result of increased international trade and deregulation of different industries, on the ability of firms to discriminate against women.
Labor market discrimination is defined as a situation in which people who provide labor market services and who are equally productive are treated unequally, and this disparate treatment is related to observable characteristics such as race, ethnicity, or gender. Early work on discrimination focused on efforts to compare outcomes of equally able individuals of different sexes, arguing that if the individuals were observably the same, except for their sex, they were likely equally productive. Some of the most compelling work applying this strategy was done by Wood, Corcoran, and Courant , who examined male and females who graduated from Michigan law school between 1972-1975. Because they compared individuals who all completed the same law school, they were able to minimize unobserved heterogeneity. In addition, because the women all made costly investments in graduate school, concerns about attachment to the labor force was less likely to be an issue. The authors found evidence of gender discrimination, even controlling for all observable characteristics. One of the limitations of this type of work, however, is that there may still be differences between the men and women that are unobservable to the economist. As a result, attributing residual differences in wages or other outcomes to discrimination may not be compelling.
More recent work has focused on the dynamic implications of discrimination. In his seminal work published in 1957, Gary Becker argued that, over the long run, product market competition would drive discrimination out of the marketplace. Becker's model as applied to the labor market can be described in relatively general terms: employers with a “taste for discrimination” will forego profits in order to indulge their desire to employ a specific type of worker. For example, employers with a taste for discrimination against women will employ fewer than the profit-maximizing number of women. Instead, they will hire a greater number of equally skilled but more highly paid men. Thus, in a perfectly competitive market, nondiscriminating employers can gain a cost advantage and ultimately drive discriminating employers out of business. Becker's model suggests that the wage gap between men and women will therefore decline as discriminators are forced to leave the market altogether.
Becker goes on to say that where markets are not perfectly competitive – that is, markets in which companies face little product market competition – discriminating employers can exist in the market indefinitely. Given the lack of transparency surrounding the practice of discrimination, Becker's theory has proved difficult to test. Recent trends in the labor market, including increased product market competition through increased international trade and product market deregulation, however, make it possible to explore some of the dynamic implications of his model.
Earlier work exploiting this dynamic relationship examined the cross-sectional relationship between degrees of product market competition and relative wages to ascertain whether or not a market or industry was discriminatory. Using cross-sectional data from New Jersey and Pennsylvania, for example, Orley Ashenfelter and Timothy Hannan  compare markets in the banking industry that are more concentrated to markets that are less concentrated. Their results show a negative relationship between market concentration and the share of female employment in each bank, a finding that is consistent with the notion of increased discrimination in concentrated markets. More recently, Hellerstein, Neumark, and Troske  use establishment level data to look at firms with market power. This study finds that, as economic theory predicts, the firms that employ a greater number of women have higher profits.
More recent work examines the dynamic implication by focusing on changes in product market competition and relates that to changes in the gender wage gap. In Black and Brainerd , the authors look at competition across industries and investigate how an increase in international trade has affected the relative wage position of women. In firms that could discriminate against women because of a noncompetitive environment, the demands of increased trade should help reduce discrimination and improve the relative wage position of women. This assumption is tested by comparing historically concentrated or noncompetitive industries that have faced increased pressure from international trade to historically competitive industries that have faced similar pressures. In theory, competitive industries should not be able to discriminate even in the absence of trade. Thus, by controlling for the general effect of trade, the effect of increased competition on the wages of women relative to men can be isolated and measured.
A notable increase in international trade is one type of shock that would tend to expose an industry to greater competition. Although increased trade should limit the ability of firms to favor particular groups of workers within an industry, it also may affect wages through other channels as well. For example, because imported goods tend to be produced by relatively low-skilled workers, increased trade may actually decrease the wages of low-skilled workers in the United States. Thus, if women are disproportionately represented among low-skilled workers, then the overall effect of increased trade may be a lowering of the relative wages of women – even if discrimination had been reduced by the increase in trade.
To control for such differences in skill level and other characteristics, the focus of the analysis is narrowed. Instead of comparing the wages of women in industries affected by trade with those of women in industries not affected by trade, the focus here is on the wage effects in industries in which the level of competition differed before the trade shock. Because industries that were already competitive would not be able to afford to discriminate before an increase in trade, increased competition should have little or no effect on an industry's ability to favor certain groups of employees. In less competitive industries, on the other hand, increased trade should reduce firms' tendency to discriminate. By comparing the effects of trade on the relative wages of women in concentrated versus competitive industries, the effects of discrimination can be isolated. The estimation nets out any factors that may have affected the gender wage gap in manufacturing industries, trade-impacted industries as a whole, or concentrated industries as a whole.
The analysis covers the 1977-94 period and is based on data drawn from the Current Population Survey (cps) for individuals aged 18 to 64 who worked full-time in the civilian sector in the year before the survey. The change in the residual gender wage gap – the dependent variable – is calculated by first regressing the log wage of all the individuals in the sample on the following variables: four categories of education level (less than a high school diploma, high school diploma, some college but no degree, and at least a bachelor's degree), age, age squared, and a nonwhite dummy variable. The residual gender wage gap is then calculated as the difference in the average residual wage for men and women at the industry level. The industry-level results are then matched to industry-level trade data from the National Bureau of Economic Research Trade Database. Trade is measured as import shares. Finally, an industry is classified as concentrated if in 1977 the four-firm concentration ratio was 0.40 or greater in the Census of Manufacturers conducted in that year.
The findings indicate that a 10-percentage point increase in import share in concentrated industries leads to a decline in the residual gender wage gap of about 6.6 percent. To understand the magnitude of this estimate, consider that the average increase in import share in concentrated industries accounts for a decline in the residual gender wage gap in manufacturing of about 0.034 log points. By contrast, during the 1977-94 period, the overall decline in the residual gender wage gap was approximately 0.14 log points. Although a positive relationship between increased trade and reduced discrimination against women is observed, the findings also show that trade as a whole is having a negative effect on the relative wages of women. This negative effect is only somewhat offset by the improvement in women's relative wages due to the decline in discrimination.
These results are tested for sensitivity in several ways. First, the same estimation strategy is applied to two other data sets – the outgoing rotation of the Current Population Survey from 1979-1994, and the 1980 and 1990 Censuses – and similar results are obtained. Second, the effects of trade on the residual gender wage gap are estimated using the metropolitan statistical area as the unit of observation. These adjustments yield relatively robust results.
Other factors could explain the findings as well. For example, one might argue that the results reflect a decline in unionization rather than a reduction in the ability of firms to discriminate. Because of higher rents, concentrated industries are likely to be more highly unionized than competitive industries, and men are more highly unionized, on average, than women. Given these differences in unionization, a decline in unionization rates over the 1977-94 period would likely reduce the gender wage gap more in concentrated industries than in competitive industries. To ensure that the results presented here do not reflect such a shift, the industry-level change in unionization is included in the regression. Despite the addition of this variable, however, the results remain the same.
To gain further evidence that the results of this study capture the effect of competition on discrimination, the employment and occupational status of women in the two samples of companies are examined. In a discriminatory environment, firms would be expected to hire fewer women and to keep them in lower positions. Looking at the change in the percentage of women employed in an industry, weak evidence was found that the employment of women increased more in concentrated industries with increased trade than in the already competitive industries with the same increase in trade. Next, the change in the percentage of managers who are women was examined, finding that this percentage increased more significantly in historically concentrated industries hit by trade. These results are consistent with the conclusion that competition reduces firms' ability to favor particular types of employees.
Other work has focused on the impact of increased competition on the ability of firms to discriminate by using the deregulation of particular industries as the shock to competition. Earlier work by Rose  examined the effect of increased competition from deregulation in the trucking industry and concluded that, prior to deregulation, there was discrimination against minorities; the evidence: after deregulation, the racial wage gap declined substantially.
More recently, focusing primarily on the gap between men and women, Black and Strahan  examine deregulation of the banking industry in the United States to see what happens to relative wages when an industry faces a shock to product market competition. The history of deregulation in the industry provides an excellent body of empirical data to study the relationship between competition and women's pay. When government regulations restricted the entry into and expansion of the banking industry, firms earned rents that could be spent in a variety of ways. After deregulation, there was a significant reduction in profits in the industry and, as a result, a decline in the ability to share these profits with “favored” workers. (See Jayaratne and Strahan 1996 for more discussion of the effect of deregulation on bank profitability.)
Black and Strahan use the deregulation of state-level restrictions on bank expansion (a unique event in recent history) to test the effect of increased competition on the labor market. Until the 1970s, banks' ability to enter new markets was constrained – only twelve states allowed unrestricted statewide branching. Over the subsequent 25 years, states gradually lifted these restrictions and banks were able to enter new markets, either by opening branches or by owning banks in multiple states. This study provides evidence suggesting that, before deregulation, rents were shared with labor and that these rents were shared disproportionately with men. This finding suggests that discrimination against women is more likely to occur in the absence of competition.
Banking deregulation provides a valuable laboratory to explore the effects of regulations restricting entry. Before deregulation, firms in the banking industry were earning rents because of the limited competition in the industry. Deregulation provided an exogenous shock to competition, which led to a decline in rents as firms were forced to improve efficiency and lower prices in order to compete more effectively.
It is important to note that deregulation in banking occurred at the state level, which is quite different from national deregulation of such industries as telecommunications and transportation. Studies that look at national deregulation cannot control for aggregate trends. By contrast, because banking deregulation occurred across states at different times, it is possible to eliminate the effects of national trends in the industry as well as The state-specific effects. For example, banking wages have been rising over the past two decades – from about $30,000 per year in 1976 to about $40,000 in 1996 (1997 dollars). Simply looking at wages before and after this period of deregulation may lead to the false conclusion that deregulation had caused the increase in wages. In fact, however, the findings of this study show that deregulation led to falling wages.
For the estimation, data from the Current Population Survey for the 1977-97 period was used, focusing again on full-time workers aged 18 to 64 who are not self-employed, working without pay, or in the military.
To test whether rents were shared with labor in the regulated environment and whether they were shared unevenly among men and women, changes in compensation and wages following deregulation are estimated. After controlling for trends in the banking industry and state-specific trends, it was found that average compensation and average wages for banking employees fell (or rose less than the trend) after states relaxed restrictions on bank branching to a statistically significant degree. The estimate is robust across different data sets and model specifications and cannot be explained by shifts in relative employment demands after deregulation. In addition, it can be shown that the decline cannot be explained by a change in observable skills, which suggests that rents were indeed shared with employees.
How much are firms able to discriminate? To test for a change in discrimination, the behavior of wages for men and women after deregulation was examined. Male wages fell about 12 percent after deregulation, while women's wages fell only 3 percent. This difference is statistically significant and suggests that before deregulation rents were shared with men more than with women. This finding is consistent with the theory that competition reduces a firm's ability to spend rents on favored groups.
As a check to this estimation, the effects of deregulation on the industry's occupational structure are examined. A discriminating employer may prefer to keep women in lower positions than their skills warrant. The findings are consistent with discrimination: the share of women holding managerial positions increased after deregulation. The share of women in managerial positions rose by about 4 percentage points or about 10 percent of the mean. These results support the theory that a lack of competition can promote costly discrimination. When competition increases, firms appear to be forced to improve the occupational status of women to cut costs. Looking within occupation groups, it appears that women's relative wages improved after deregulation in part because their relative wages within their occupation improved and in part because they moved into higher skilled occupations.
A key concern is that these findings may simply reveal a decline in the demand for labor. To test for shifts in labor demand, therefore, the change in the percentage of workers in the banking industry in each state was examined. The analysis finds no significant decline in this percentage after deregulation, which suggests that it is not simply a shift in labor demand. In sum, it appears that an environment of increased product market competition reduces employers' ability to practice discrimination.
It is clear that the relative status of women in the labor market has improved, and this improvement is both evident over time and across countries. Understanding this improvement can be difficult; while some of the change can be attributed to changes in the characteristics of women, such as increased education and labor market experience, other components are more elusive, such as changes in the discriminatory environment that they face. Recent work on discrimination has demonstrated that there is a clear link between the increased product market competition we see across sectors and the level or amount of discrimination faced by women.
Women are making economic progress, as has been documented in this paper. But we don't yet have all the answers. How much discrimination remains in the market? And why, in some cases such as college enrollment, have women not only caught up to men but actually surpassed them? I leave these questions to future research.
Sandra E. Black
[ *] Assistant Professor of Economics at the University of California, Los Angeles, a faculty research fellow at the nber, and a research associate at iza. She can be contacted at sblack@ econ. ucla. edu or by mail at the Department of Economics, 8283 Bunche Hall, ucla, Los Angeles, CA 90095.
[ 1] Note that bank-level data also are used in the first part of the analysis to demonstrate that, prior to deregulation, rents were shared with workers.