The Labor Market Effects of Sex and Race Discrimination Laws
Neumark, David, Stock, Wendy A., Economic Inquiry
Public policies to narrow the gaps in labor market outcomes between men and women and between whites and minorities have a long and controversial history. Two pieces of federal legislation stand out as perhaps the most significant such policies. The first was the Equal Pay Act of 1963, which requires equal pay for equal work, noting some exceptions but explicitly prohibiting a worker's sex as one of them. The second was Title VII of the Civil Rights Act of 1964, making it illegal to discriminate in hiring, discharge, compensation, and so on, on the basis of race, color, religion, sex, or national origin. Two other important landmarks in the evolution of these policies at the federal level were amendments to Title VII embodied in the Equal Employment Opportunity Act (EEOA) of 1972, which expanded coverage and increased the enforcement powers of the Equal Employment Opportunity Commission (EEOC), and Executive Orders 10925 (1961), 11246 (1965), and 11375 (1967), which laid the groundwork for affirmative action, although the phrase has its origins in Title VII. (1)
There are three broad controversies regarding these public policies. The first concerns the need for any government attack on labor market discrimination, hinging on questions of whether the observed group differences in labor market outcomes reflect discrimination and whether competition in labor markets and product markets will undermine discriminatory behavior. (2) The second concerns the fairness and social efficiency of affirmative action, which many, such as Steele (1990) and Carter (1991), regard as distinct from nondiscrimination policies in advocating preferential treatment of particular groups. (3) The third, with which this article is concerned, is the effectiveness of these policies, asking in particular whether antidiscrimination policies contributed to relative improvements in labor market outcomes for women and minorities (and presumably continue to contribute if discrimination persists). Failure to find evidence of these effects can reflect either ineffectiveness of the antidiscrimination laws, or no prior discrimination.
We focus much of our attention on the effects of laws prohibiting sex discrimination in pay, unlike most of the previous research on the impact of antidiscrimination legislation, which focuses on the effects of race discrimination laws on race differences in earnings. Much of the earlier work tries to infer the effects of federal legislation by asking whether, concurrent with the passage of federal antidiscrimination legislation, there was a jump or acceleration in, for example, the relative promotion rates of women or the relative earnings of blacks. The fundamental problem with such time-series analyses of the impact of federal antidiscrimination laws is that the laws have nearly universal applicability, which prevents identification of an appropriate comparison group that can be used to control for changes in the relative outcomes under study that are unrelated to the policy innovation. For example, if the black-white wage gap was narrowing prior to the passage of Title VII of the Civil Rights Act, then testing whether Title VII narrowed the gap requires a comparison of changes in the black-white gap for workers covered by Title VII and workers not covered in the same period. (4) Researchers have of course considered other ways to bring complementary evidence to bear, including efforts to identify control groups and use of auxiliary data.
This article takes an alternative approach to the problem of inferring the effects of laws prohibiting sex and race discrimination to those taken in most previous research. In particular, prior to the enactment of the federal legislation, many states enacted similar laws or practices barring discrimination in wages (for women) and employment (for blacks, although those laws covered wage discrimination as well). Because these laws or practices were passed at different times in different states, a more natural control group is provided. Specifically, the quasi-experimental design afforded by the variation across time and over states allows us to assess the impact of state antidiscrimination statutes by constructing comparisons using data for the same time span from states that did not enact such statutes. (5) Although this approach has some limitations, at a minimum our empirical strategy provides important complementary evidence to that in the existing literature on the impact of federal antidiscrimination laws. A recent survey of research on the effects of gender and race in the labor market by Altonji and Blank (1999) concurs, summarizing its review of existing evidence on the effects of federal antidiscrimination policy as follows:
Despite major public and private resources devoted to anti-discrimination policy, the research literature on the results of these efforts is sparse. While we recognize the difficulties of studying nationally enacted legislation, in many cases there are differences over time or across regions in the implementation of such legislation, or there is variation in related state-specific legislation. Such research ... is likely to provide useful information, particularly in a world where existing anti-discrimination measures in ... the labor market are at the center of a major public debate about the appropriate response to ongoing racial differentials. (p. 3250)
II. EXISTING RESEARCH
The body of research on the impact of federal antidiscrimination policy on sex differences in labor market outcomes is quite small. Eberts and Stone (1985) use panel data to examine relative rates of promotion to administrative positions of male and female public school teachers before and after the EEOA. They find declining evidence of discrimination in promotions of teachers in two states in the latter part of the 1970s and conclude that the EEOA contributed to that decline. But as just noted, attributing such changes to antidiscrimination laws rather than concurrent changes in labor markets and labor force behavior of males and females is tenuous.
Beller (1979) improves on this identification strategy by using information on regional variation in policy enforcement activity. She estimates models for earnings of men and women using Current Population Survey data for 1974, 1971, and 1967, incorporating measures of Title VII investigations and settlements by region (large states and state groups) but not including a time trend or year dummy variables. She finds relatively weak evidence that these measures reduced the sex wage gap prior to the EEOA (in 1972), but stronger evidence of this effect after the passage of the EEOA. O'Neill (1985) studies the evolution of the sex gap in wages but without an explicit focus on policy effects. (6)
There is a much more extensive set of time-series studies of the impact of legislation prohibiting race discrimination, examining evidence regarding alternative explanations of black economic progress in the late 1960s. The primary question addressed in the time-series studies is, as Heckman (1990) puts it, "Does continuous or discontinuous change characterize the recent economic history of black Americans?" (p. 242). That is, is there a discontinuity in the relative progress of blacks that is most consistent with an important role for federal antidiscrimination efforts? Heckman takes a strong stand regarding the positive impact of federal policy (based on evidence in Donohue and Heckman 1991, and a review of the earlier evidence), arguing that "there is ample evidence of discontinuous change in the improvement of black status during the crucial period 1965-1975" (1990, p. 242). (7) A principal alternative view is that black economic progress simply reflects longer term trends, perhaps obscured in some periods (and hence giving the impression of more rapid change in the 1960s and early 1970s) because of other changes. The latter view is put forth most forcefully by Smith and Welch (1989), who conclude that "slowly evolving historical forces ... education and migration--were the primary determinants of long-term black economic improvement. At best affirmative action has marginally altered black wage gains around this long-term trend (p. 519). (8) A central reason for this view of the evidence is that relative black economic outcomes had been improving earlier, although this may have been partly masked by other changes. As an example, Margo (1995) and Goldin and Margo (1992) have documented the Great Compression of general wage inequality as well as black-white earnings differences from 1940 to 1950, (9) and, as Margo asks, if antidiscrimination legislation played a strong role in racial wage convergence in the 1960s, "how did black workers manage such impressive relative wage gains in the 1940s, well before the modern civil rights movement bore its fruit?" (1995, p. 470).
To illustrate these facts, Table 1 reports estimates of specifications using census data that mimic the earlier time-series studies, simply inferring the effects of anti-discrimination legislation from changes in relative earnings and employment over time. (10) We estimate individual-level equations for employment and for log earnings, including dummy variables for race, year, and state, as well as control variables listed in the notes to the table. We augment the specifications to include interactions between year dummy variables and a dummy variable for blacks or women. The year dummy variables are defined to equal 1 if an observation is in or after a particular census year (e.g., year [greater than or equal to] 1960), so that each of these interactions measures the change from the previous decade. (11) The identification strategy for the time-series analysis can be interpreted as comparing the coefficient estimates for race or sex differences in 1970 to those for earlier decades, with a discrete jump or an acceleration providing evidence of an effect of the federal legislation, in the spirit of the discontinuous change criterion. As we show in the bottom panel of columns (2) and (4) of Table I, in the census data that we use in this article we find the same evidence of earnings gains for blacks relative to whites between 1960 and 1970. But we also see the sharper relative earnings improvement for blacks from 1940 to 1950. The table also shows gains in the relative employment of both black and white women from 1950 to 1960 as well as 1960 to 1970.
The evidence in Table 1 makes two points. First, for the most part relative economic outcomes for women and blacks improved from 1960 to 1970, consistent with beneficial effects of federal antidiscrimination policies. On the other hand, changes were in some cases occurring in earlier decades, and the earlier changes were sometimes larger. Together, these findings make it difficult to draw strong conclusions regarding the effects of federal antidiscrimination policies enacted in the 1960s. Even if the changes in all other decades were 0, and in the expected direction from 1960 to 1970, alternative explanations other than effects of these policies would be possible, although less plausible. But the estimates reported in Table 1 are far from that scenario, emphasizing that it is difficult to draw inferences regarding the effects of antidiscrimination policies solely from changes over time in relative outcomes for protected groups, a conclusion that helps to motivate our analysis of state-level antidiscrimination laws.
In the face of the problem of identifying a comparison group that can be used to control for changes in the outcomes under study unrelated to the policy innovation, researchers have used additional information and/or alternative empirical strategies to identify the policy effects, based on regional differences in antidiscrimination efforts (e.g., Beller 1979; Donohue and Heckman 1991; Heckman and Payner 1989) and differences between businesses covered and not covered by antidiscrimination laws (e.g., Chay 1998; Carrington et al. 2000; Hahn et al. 1999; Leonard 1984). The studies in the second group are limited compared to the overall question regarding the effectiveness of antidiscrimination efforts, because they tend to focus on more narrowly tailored evidence, such as the effects concentrated at particular employer-size cutoffs. The evidence we present, based on variation across states and time in the introduction of antidiscrimination legislation, is of a more aggregated nature and in that sense closer in spirit to the earlier time-series studies in providing estimates of overall policy effects, while also building on the general empirical strategy of the newer research that develops better experiments for evaluating the effects of antidiscrimination legislation.
III. SEX AND RACE DISCRIMINATION LAWS
This article exploits cross-state variation in antidiscrimination legislation at the state level. Thus a major part of the research project was assembling information on this legislation with regard to both sex and race discrimination, as well as related legislation potentially impacting women. Information on laws regarding sex discrimination is summarized in Table 2. As the table shows, in almost every case through the 1960s these were equal pay laws and did not explicitly refer to discrimination in hiring, discharge, and so on. The only exceptions are a voluntary Fair Employment Practices Act (FEPA) provision in Oregon as of 1950, and then, as of 1970, the extension of FEPAs to refer to sex in numerous states. In contrast, mandatory FEPAs, which were widely adopted by states with respect to race, establish two types of behavior as violations of the acts: refusing employment or discharging a nonwhite because of that person's race; and discriminating against nonwhites in terms or conditions of employment, including compensation. (12)
The state laws for women raise a couple of issues. First, given that the state laws cover wage discrimination only, the federal laws enacted in the 1960s are fundamentally different because they also incorporated nondiscrimination in hiring and discharge. As a consequence, we focus only on the impact of state equal pay laws through 1960. That is, there is no way to think of the passage of federal laws in the 1960s as the "extension" of similar laws to the remaining states, given the greater breadth of the federal laws regarding sex discrimination. (13)
A second issue is the potentially confounding effects of protective laws restricting women's work. At the end of the nineteenth and beginning of the twentieth century, most states adopted laws that in some way or other restricted women's work, most commonly by specifying maximum hours or prohibiting employment in particular occupations or at night. A detailed compendium of these laws is provided in Table 3. The types of laws covered include prohibitions on night work, prohibitions on employment of women in certain occupations and industries, and maximum hours laws. As the last column of the table shows, state attorneys general or other state bodies sometimes issued opinions in the 1960s noting that these laws were inconsistent with Title VII of the Civil Rights Act, so that such laws were by and large nonbinding by 1970. Nonetheless, if these laws were changing over the 1940-60 period, any effects of such changes could be difficult to sort out from changes in antidiscrimination laws. As the table shows, however, with the exception of Delaware, there was no activity regarding these laws in this period. (14) As the table also documents, only following the Civil Rights Act of 1964 did the dismantling of protective legislation occur. Thus, it turns out that we need not be concerned about confounding the effects of state antidiscrimination laws with effects of changes in protective laws.
Finally, information on race discrimination laws is tabulated in Table 4. The most common type of legislation is the enactment of a mandatory FEPA (e.g., Alaska, California, and Massachusetts). However, Table 4 also reveals that there is heterogeneity in the FEPAs, including variation in exemptions for employers below certain size thresholds (e.g., Indiana, Minnesota, and Pennsylvania), differences in enforcement mechanisms (e.g., New Jersey, New York, and Wisconsin), and whether compliance is mandatory or voluntary (e.g., Colorado, Kansas, and Oregon).
Table 5 summarizes our coding of the sex and race discrimination laws. Because we study the effects of these laws using data from the decennial census, we only show the laws in effect at the end of each decade. In the empirical work, we do not pay attention to all of the variation in state legislation documented in Tables 2 and 4, in part because we do not have the requisite information (such as employer size). Table 5 does distinguish mandatory FEPAs (MRD), voluntary FEPAs (VRD or VSD), and race discrimination laws that target only wages (MRDW), although the results reported in the tables are from specifications limited to mandatory laws. (15) The table also distinguishes laws passed in years ending in 9, which are the years covered by the Census earnings questions; in the empirical analysis, we allow for such laws to have potentially weaker impacts, because they were in effect only part of the year. (16) Finally, because the federal law kicks in during the 1960s and supercedes the state law, Table 5 only reports the coding through 1960. (17)
The sex discrimination laws we study are limited to equal pay laws that prohibit pay discrimination. A question arises as to how interesting evidence on the effects of such laws is--from a policy perspective--given that federal policy essentially enacted equal pay legislation and prohibitions of employment discrimination simultaneously. In our view, there are two reasons that an analysis of equal pay laws is of considerable policy interest. First, the ongoing debate over affirmative action may eventually weaken laws barring employment discrimination in hiring. Although the policy debate is focused on affirmative action per se, numerical guidelines also play a critical role in antidiscrimination enforcement efforts, as noted by Bloch (1994). Although we do not anticipate a return to a world where sex (or race) discrimination in labor markets is regulated only by equal pay laws, we can envision one in which these laws become a relatively much more important part of the arsenal of antidiscrimination efforts, making the independent effects of equal pay laws of interest.
Second, there is considerable interest in the effects of equal pay laws per se because of skepticism about whether the effects of prohibitions on discrimination in hiring played a major role, particularly in black economic progress in the 1960s, and conversely because of a sense that equal pay laws are more enforceable. For example, although arguing that the overall thrust of federal antidiscrimination policy helped spur black economic progress in the United States beginning in the mid-1960s, Donohue and Heckman (1991) argue that this progress was due to much more than efforts to eradicate hiring discrimination via direct EEOC enforcement (and affirmative action), and also depended on efforts to combat wage discrimination, as well as on broader federal efforts extending beyond the labor market. Others, such as Bloch (1994), have noted that antidiscrimination laws targeting employment discrimination may be more likely to be effective against discrimination in discharge than in hiring, and conversely that equal pay laws may be quite effective because of the combination of a readily identifiable and potentially large class and potentially high damages.
A final question concerns whether state laws have any "teeth." With regard to state laws requiring equal pay for women, although we have been unable to uncover much direct empirical evidence on enforcement activities, resolution of disputes, and so on, the historical record does point toward enforcement of these laws and perceptions of them having some impact. According to a U.S. Department of Labor report discussing state equal pay laws, "The labor administrators of the respective States are generally responsible for enforcement of State equal pay laws. However, experience in the equal pay States indicates that once a law is enacted, with its subsequent publicity which serves an educational purpose, most employers comply voluntarily with its provisions" (U.S. Bureau of Labor Standards, 1967, p. 249). (18) Although this may just reflect wishful thinking, the report does cite two cases--in Michigan and in California--in which female employees who had suffered from wage discrimination were given financial awards under their states' equal pay laws.
In addition, the record from a 1963 U.S. Senate hearing on the Equal Pay Act provides a number of different types of evidence on the relevance of earlier state equal pay laws (Committee on Labor and Public Welfare 1963). First, a digest of such laws submitted by the Women's Bureau of the U.S. Department of Labor (1970, pp. 43-54) lists record-keeping requirements (such as wages, salaries, occupation, job classification, etc., of each worker), possible court actions (generally employee lawsuits, possibly as a class), and penalties (for example, back wages plus fines or even an additional amount equal to back wages) and indicates which administrator in each state is responsible for enforcement. Second, the record from the hearing also lists selected court cases involving state equal pay laws. It is unclear on what basis the particular cases listed were chosen, but the evidence of litigation surrounding these lawsuits indicates that they had some teeth. Third, the Women's Bureau also submitted evidence from inquiries sent to the states with equal pay laws regarding the use of administrative or informal hearings to resolve equal pay disputes, and of the 18 that responded, 11 verified that they held some hearings in administering their statutes. Many of these reported that some actions were taken, and four are quoted as attesting to the effectiveness of their compliance efforts (Women's Bureau 1970, pp. 63-64). Fourth, the record includes testimony from William Miller, representing the U.S. Chamber of Commerce, suggesting that the chamber was at least concerned that these laws have some impact: "More and more women are beginning to realize that the more you legislate specifically with respect to women, the more you hurt their job opportunities" (Women's Bureau 1970, p. 71). Similarly, former Labor Secretary Maurice Tobin is quoted indirectly as saying that "employers in States which have equal pay laws and consequently higher wage standards are often put at a competitive disadvantage with those in States where there is no legal barrier to unfair exploitation of women workers by means of discriminatory wage cutting practices" (Women's Bureau 1970, pp. 91-92). At the same time, the record also suggests some concerns with the effectiveness of the state laws. James B. Carey, representing the AFL-CIO, is quoted as saying: "Twenty-two states now have equal pay laws of varying effectiveness. Unfortunately, too many are limited in scope, while others have few enforcement teeth" (Women's Bureau 1970, p. 83).
Turning to race discrimination laws, there is somewhat more information suggesting that these laws had some actual impact. Mandatory state FEPAs can be viewed as operating in a similar fashion to Title VII of the Civil Rights Act. (19) Landes (1968) presents some descriptive information on violations "cleared up" in New York, New Jersey, Connecticut, and Massachusetts in the period between 1945 and 1961. The number of cases ranges from about 500 to 700 for the three smaller states, to nearly 3,300 for New York. These cases in turn are only a subset of complaints for which state commissions found probable cause to support the complaint. Lockard (1968) presents data on the number of FEPA cases closed over the 1940s, 1950s, and 1960s in 10 states, with the numbers ranging from about 150 to 6,000; he also reports that in 64% of cases no probable cause was found. Similarly, a report written by the New York State Commission against Discrimination (1958) details efforts to counter discrimination in the hotel industry in New York City, including information on complaints received and the settling of those complaints that were sustained (i.e., judged as valid) through conference and conciliation. (20)
Addressing the question of effectiveness, Lockard (1968) cites a study by the New Jersey Division against Discrimination showing that among 54 firms that had been involved in discrimination cases that resulted in adjustments in behavior, total employment was up 22%, but minority employment was up 107%. Undermining the claims of effectiveness, however, Lockard argues that budgets for enforcement were low (although providing no basis for comparison), and notes cases where top administrative positions were left empty (1968, pp. 82-94). Of course, one could argue that the EEOC has similarly experienced periods of weak enforcement of antidiscrimination laws at the federal level.
Ultimately, we cannot definitively establish the level of enforcement of state antidiscrimination efforts, especially in comparison with later federal efforts. But it appears clear that there was indeed some antidiscrimination enforcement at the state level prior to the federal legislation, based on …
Questia, a part of Gale, Cengage Learning. www.questia.com
Publication information: Article title: The Labor Market Effects of Sex and Race Discrimination Laws. Contributors: Neumark, David - Author, Stock, Wendy A. - Author. Journal title: Economic Inquiry. Volume: 44. Issue: 3 Publication date: July 2006. Page number: 385+. © 2003 Western Economic Association International. COPYRIGHT 2006 Gale Group.
This material is protected by copyright and, with the exception of fair use, may not be further copied, distributed or transmitted in any form or by any means.