Protect Us, Lord, from Richard Epstein

Jonah Gelbach & Lesley Wexler & Jonathan Klick

Posted in ,

This article is a response to an earlier posted piece by Richard Epstein: Protect Us, Lord, from Title VII: A Response to Gelbach, Klick, and Wexler.

We thank Richard Epstein for commenting on our online Article. He brings a unique perspective to the field of employment discrimination and pushes other scholars to think carefully about their additions and amendments to the existing regulatory landscape. We have some fundamental disagreements with Epstein, but a careful reading of our full Article easily addresses most of his criticisms. By way of quick summary, let us remind the reader of our original premise:

What we term “passive discrimination” involves the employer’s use of wage and benefits packages that exploit observed, systematic group-level preference heterogeneity to induce workers to sort themselves ex ante such that members of a disfavored group view the job opportunity as being less attractive than members of other groups. . . . We note that some employers hold a discriminatory intent when designing such terms and conditions, while others might be simply unaware or neutral as to when their design of such packages will induce segregation based on membership in a disfavored group.1

While we three disagree as to Title VII’s ultimate desirability and utility, Epstein unfairly characterizes our original Article as unabashed Title VII expansionism. A look at the Article shows our careful acknowledgement of Title VII’s limitations and our embrace of creative, nonregulatory solutions to the problem of passive discrimination.

Protecting Choice while Avoiding Liability

Epstein’s hostility to our Article’s recommendations is misplaced. He mistakenly contends that we discourage menu options and argues,

One of the most common practices for fringe benefits today is for firms to offer workers an allowance that they can use to make purchases from a menu of items, in exchange for a reduction in their base pay. . . . This ability to make the optimal choice thus counts as an implicit wage boost for all employers, regardless of race or sex. . . . [T]he employer practice should be sheltered by the doctrine announced in EEOC v Sears, Roebuck & Co, because we have the most explicit declaration of unconstrained worker preferences that we can imagine. . . . The widespread use of these menu options is not consistent with employers engaging in covert forms of discrimination.  The menu strategy is an effective way to attract a diverse workforce because it does not shoehorn the benefits package into a one-size-fits-all straightjacket.  GKW should praise employer ingenuity for maximizing the welfare of their employees under competition, not seek ways to expose them to additional liabilities.3 For instance, we state, “For most of the problems identified in this Article, figuring out ways to enhance employee choice should help reduce benefits discrimination for those with atypical preferences. . . . Such choice allows employees to select a package that best matches their expectations and preferences.”4 We affirmatively encourage specific possibilities as compensation menus for sales positions as well as urge the EEOC to develop gold standard programs to get employers to think creatively about the provision of equally valued fringe-benefit packages.5 An employer who provides such choice in fringe benefits or compensation packages would most assuredly not be subject to liability under even our most far-reaching proposals to amend Title VII. Thus, Epstein is clearly mistaken in his assertion that “every case will have either formal or impact discrimination.6 Though a firm that “picked a mandatory insurance benefit which it then deliberately mispriced to drive women applicants from the roost”7 would be held liable under our expanded notion of Title VII, a firm that offers a variety of health insurance policies, only some of which are priced to account for women’s lower risks of death, would not face liability.

Similarly, we think Epstein misunderstands our argument when he notes, “The widespread use of [ ] menu options is not consistent with employers engaging in covert forms of discrimination,” and that we “should praise employer ingenuity for maximizing the welfare of its employees under competition, not seek ways to expose employers to additional liabilities.”8 As we explain more fully below, a firm offering compensation menus, where each item on the menu imposes the same cost on the firm, cannot be engaging in passive discrimination. Allowing workers to choose compensation form vitiates any screening function of compensation packages. Thus, when Epstein writes that “[t]he widespread use of these menu options is not consistent with employers engaging in covert forms of discrimination,” he has our—and our model’s—full-throated agreement. Rather than disproving our argument, this admonition of Epstein’s is one of our model’s primary conclusions. We are thus pleased to agree with Epstein on this crucial point of ours.

Passive Discrimination: Pervasive or Possible?

While Epstein suggests that most employers avoid passive discrimination by offering menus, we make no claim that intentional or even unintentional passive discrimination is widespread. Epstein further notes that he is “not aware of any evidence that points to a conclusion that any affirmative action employer has ever engaged in this tactic, either deliberately or inadvertently.” We are also unaware of any intentional affirmative action passive discrimination,9 which is consistent with his point that aggressive recruitment of minorities is unlikely to be punished by Title VII and thus minority-seeking employers need not pursue such a strategy. Stipulating for discussion’s sake only that courts treat affirmative action benevolently, Epstein’s argument proves too much: passive discrimination is likely to be a one-way ratchet for animus-based discriminators. This observation, if true, is consistent with what we argue in our original Article that “recent class actions suggest much of legal academia and the public underestimate the prevalence of basic animus- or stereotype-driven discrimination,” and thus, it “should be unsurprising if litigation-savvy employers might deliberately craft compensation structures and packages to exclude certain types of workers.”10 In other words, academics like Epstein may be too sanguine about the existence of animus- and stereotype-based discrimination. Passive discrimination is just one particularly litigation-savvy strategy of many possible options to discourage disfavored individuals from joining or staying in a particular workplace.

While we are not aware of sufficiently detailed publicly available datasets that would allow us to test for the existence of passive discrimination, 11 we do think that our model provides insight into the selection effects (by race, gender, and so on) that exist across industries that have been consistently documented in the labor economics literature.  While labor economists generally take these selection effects as given, our framework holds the promise of explaining why such selection effects arise.

We do not definitively identify any examples of intentional passive discrimination, because without conducting a rigorous case study, we cannot confidently distinguish between intentional and unintentional passive discrimination. Of course, as we note in our Article, this difficulty is part of the problem. Without a smoking gun, scholars and courts will find differentiating between intentional and unintentional passive discrimination quite cumbersome. Yet our Article identifies several examples that likely support our hypothesis. For instance, one example we describe in the Article is Chick-fil-A and its Sundays-off policies. Chick-fil-A may have multiple motives for its Sundays-off policies, but given its frequent embroilment in Title VII suits for religious discrimination,12 we would be unsurprised if it used its Sundays-off policies as part of a larger strategy to screen out non-Christian franchise owners and employees. Other Chick-fil-A hiring practices, such as a year-long vetting process with multiple interviews for key positions, support this hypothesis.13 says the company works hard to select people like Yokum, who “fit.” “We want operators who support the values here,” [he] says. Emily Schmall, The Cult of Chick-fil-A (cited in note 15).]

Theoretical and Empirical Economic Issues

We also wish to address a variety of economic points that Epstein raises. These include Epstein’s discussion of pensions and social security, his recommendation to repeal employment discrimination laws but for a very narrow monopoly exception,14ust repeal the employment discrimination laws in their entirety, except as they apply to monopoly situations, of which there are virtually none in private unregulated markets.” Epstein, Protect Us, Lord, from Title VII (cited in note 3).] and the contention that the economic viability of passive discrimination requires identical compensation demands across races and homogeneous preferences within race.15 We also seek to deal with his related claims that we fail to provide evidence suggesting “that the position of African-American workers lags behind that of whites, controlling for the usual key differences in education and work experience and the like,”16 and Epstein’s contention about the “persistent finding[] [ ] that wage gaps between blacks and whites prove significant for men, but not for women.”17 Epstein claims that “[t]he explanations for that disparity are likely to prove complex, but invidious discrimination on racial grounds does not look to be one of them.”18

Before we address these issues specifically, we will review some basic methodological facts concerning the theoretical stability of labor market discrimination. Consider first the case of an employer operating in a perfectly competitive labor market and a perfectly competitive output market with free entry. Perfect competition in the labor market means that workers must be paid at least their marginal revenue products—the sales value to employers of the output that workers produce. An employer who wishes to ensure that she can hire only favored types of workers will have to pay a premium, marking up the favored employees’ wages above marginal revenue product. In a competitive output market, such an employer will necessarily lose money and ultimately go out of business.19 This is the standard economic argument against the long-run necessity of antidiscrimination laws: competitive markets chase out bigots.

Traditionally, economists have recognized three important counterexamples in which discrimination in long run equilibrium is possible given laissez faire policies. In these cases, antidiscrimination laws might be necessary to eliminate discrimination. Epstein himself raises the first case, one in which employers have labor market power that is somehow protected from another employer’s entry into the labor market. In this case, employers can pick and choose the workers they want, at least up to a point. In such monopsony cases, workers’ marginal revenue product is above the wage.20 Moreover, workers’ only recourse against a monopsonist who refuses to hire them is to work in some other industry, which presumably entails a loss of welfare. Even when the monopsonist must sell on a competitive output market, it earns economic rents from its labor market power. One way to spend some of those rents is to discriminate in choosing workers, and no amount of output-side competition will eliminate this option.

The second case in which equilibrium discrimination is possible under laissez faire occurs when employers have pricing power on their output markets, regardless of whether they have pricing power in the labor market. If entry is limited, then economic rents will not be totally competed away. Once again, the employer can spend some of her economic rents on wage premia that induce the favored mix of workers.

The third case in which equilibrium discrimination is possible under laissez faire involves not market power, but consumer preferences. Suppose consumers prefer to buy from firms that discriminate against disfavored workers. Given equal marginal productivity of workers, any firm that employs disfavored workers must operate at a competitive disadvantage vis-à-vis firms that do not. As a result, nondiscriminating firms will go broke, and only discriminating employers will remain in business. If consumers are willing to pay to avoid disfavored workers, no purely competitive mechanism exists to stop them from doing so.

Our model’s methodological contribution is to establish a fourth counterexample. We show that discrimination is possible, in the form of workplace segregation, when (a) all markets are perfectly competitive, and (b) consumer preferences are nondiscriminatory. What drives our result is the correlation between worker type and worker preferences for amenities. This correlation allows firms to design limited compensation packages that will lead to equilibrium segregation of the workplace, and this is the practice that we have termed passive discrimination. Our main economic methodological contribution is thus to point out a gap in the previous understanding of the conditions that allow persistence of employment discrimination.

A.     Pensions and Social Security

Epstein suggests that pension systems are unlikely to function as screening devices, since “for many jobs the pension element has little or no traction insofar as the . . . social security system covers most of the pension obligation.”21 This is an unusual characterization of Social Security’s role—one we believe contradicts decades of public policy. Consider this discussion, from the Social Security Administration Historian’s Office:23 ruling that an employer may not deduct more from women’s pay to cover their pensions even though, as an actuarial matter, women as a class are likely to draw on pension benefits longer. We do not disagree with Epstein’s argument that women’s longer longevity makes such practices economically reasonable. But we also do not see the relevance of this conclusion, either to the possible existence of passive discrimination or to our policy suggestions.

B.     Epstein’s market Power Point Proves Too Much

Epstein supports antidiscrimination employment laws only when perfect monopsony power exists. As our discussion above suggests, the obvious economic justification for this position is that no competitive forces will prevent a perfect monopsonist from sustained use of discrimination. The discussion above notes that equilibrium discrimination can occur when there is output market monopoly, or competitive markets with bigoted consumers. The economic argument behind our model shows that the combination of bigoted employers and type-correlated worker preferences is a fourth case allowing equilibrium discrimination, even in competitive markets. Epstein’s position that only the first of these four cases justifies antidiscrimination laws cannot be supported on economic grounds. If antidiscrimination laws make sense for any of these cases, we believe they make sense for the others.24

C.     Epstein Contradicts Himself on Reservation Wages

Epstein also contends that intentional and unintentional passive discrimination will not appear if disfavored groups such as African-Americans have lower reservation wages. But for employers operating as price takers on both the labor and output markets, nothing substantive in our conclusions hinges on this question. Epstein’s suggestion to the contrary relies on the surprising, if implicit, assumption that the labor market does not operate to drive wages toward marginal revenue product. Epstein writes:

Nor is there any reason to think that GKW’s strategy for racial discrimination is likely to yield any forbidden fruits if put into play.  In order for this to work, there has to be an assumption that the compensation demands across races are identical for both African-American and white workers.  But suppose that they are not, and that on average the African-American workers have lower reservation wages than white workers, perhaps because of the discrimination in the general market.  At this point, the pension-heavy strategy may well reduce the attractiveness of the compensation package somewhat. But that result is consistent with a loss in consumer surplus, and does not require us to assume that there will be much of a change in rate at which African-Americans accept job offers.25

Epstein’s suggestion here is that lower reservation wages among African-Americans will prevent firms from screening them out by converting some cash compensation to fringe benefits that African-Americans value less than whites. On this argument, the effect of such screening would be to reduce the surplus received by African-American workers, but not to change the mix of employees hired.

But Epstein’s argument hinges critically on the implicit assumption that competition in the labor market will fail to show up for work. According to the basic logic of competition, any employer who pays African-Americans in a way that does not, as Epstein phrases it, “maximiz[e] the welfare of its employees under competition,”26 will soon be searching for a new employee willing to forgo welfare maximization. A worker paid less than the value of her marginal product by her present employer will make an attractive hire from some other firm’s perspective. Therefore, her wage will be bid up to her marginal revenue product.27 Nowhere in this discussion do reservation wages play a role, except insofar as they are so high that a worker would choose to exit employment altogether—and that is the opposite of Epstein’s hypothesis.

Thus, Epstein’s point requires one to suspend belief in the forces of market competition. This is a curious position, given Epstein’s reliance on the beneficial forces of market competition in his campaign to “repeal the employment discrimination laws in their entirety, except as they apply to monopoly situations.”
28 In sum, Epstein’s suggestion that anything important hinges on the question of whether African-Americans have lower reservation wages appears to stand in significant tension with his later suggestion that buyer-side labor market power is nonexistent in private, unregulated labor markets.

Epstein also emphasizes preference variation within worker type: “We know in addition that even if there is a variation in preferences across groups, there is also a variation in preferences within groups.”29 While this statement is surely correct, it is just as surely beside the point. All that is required for the economic viability of passive discrimination is that amenity preferences be correlated with worker type—not that they be uniform within type. At the cost of much more notation and more equations, we could easily have written up a much more general version of our model that allows for any finite number of groups, each having its own nondegenerate preference distribution. Within-group preference uniformity is simply a modeling assumption that keeps the analysis wieldy, which is good practice in economic exposition.

D.     Empirical Evidence on the Black-White Gap in Labor Market Positions

Epstein criticizes us on the grounds that we

do not offer any aggregate statistics that indicate that the position of African-American workers lags behind that of whites, controlling for the  usual key differences in education and work experience and the like. That literature is of course enormously complex, and among its most persistent findings are that wage gaps between blacks and whites prove significant for men, but not for women. The explanations for that disparity are likely to prove complex, but invidious discrimination on racial grounds does not look to be one of them.30

Leaving aside the problems with the study Epstein cites,31 he rightly notes that we did not offer detailed empirical evidence on the relative position of black and white workers. Of course, such evidence was neither the focus of our argument, nor necessary to any of our conclusions. However, we now happily take this opportunity to briefly discuss some interesting evidence from this literature.

The econometric study likely most friendly to Epstein’s conclusion argues emphatically against controlling for what Epstein calls “key differences in education and work experience and the like.” In that study,32 Derek Neal and William Johnson use a sample of blacks and whites born between 1957 and 1964. They find that essentially all of the black-white wage gap among women, and about 70 percent of the gap among men, disappears when one (a) controls for scores achieved on a test administered before most sample members had entered the labor market, but (b) does not control for Epstein’s “key differences in education and work experience and the like.”33

While Neal and Johnson offer an economic argument in favor of this approach to estimating the black-white wage gap, labor economists disagree on this point. For example, after extensive theoretical and empirical investigation, including a reanalysis of Neal and Johnson’s data, Kevin Lang and Michael Manove34 conclude that “there are good grounds for believing that at least some of the black-white wage differential reflects differential treatment in the labor market.”35 Consider also Joseph G. Altonji and Rebecca M. Blank’s widely cited Handbook of Labor Economics chapter.36 Among their “key conclusions” concerning differentials in wages and labor force participation, Altonji and Blank offer that

[e]ven controlling for occupation, industry, and job characteristics, there remain significant differentials between white males and other workers. Some of this may be due to incompletely specified models. . . . Some of it almost surely represents ongoing constraints in the labor market for women and minorities.37

Having discussed such evidence, we note two reasons why Epstein’s concentration on wage gaps seems misplaced. First, in our model’s segregated equilibria, favored workers receive some of their compensation in the form of fringe benefits, whereas disfavored workers receive all of their compensation in cash. Competition forces firms to spend the same amount to employ each type of worker. It follows that disfavored workers will be paid more in cash than equally productive favored workers. A more powerful empirical critique of our model than Epstein’s would thus seem to be that in the real world, employers generally do not pay African-Americans more cash compensation than they pay whites. However, a valid empirical test of our model would require overcoming two empirical hurdles: getting data on worker-specific marginal productivity; and dealing appropriately with difficult econometric issues related to workers’ job choices, which is the key mechanism through which passive discrimination would operate. We leave the interesting problem of solving these nontrivial econometric problems for future research.

More important for our Article, Epstein’s focus on wage gaps is misplaced, as our primary focus is on workplace segregation, not on wage discrimination. For evidence on workplace segregation, consider recent research by Judith Hellerstein and David Neumark,38 who study “segregation in the labor market—that is, the extent to which members of different groups tend to work with coworkers who are more like themselves than would be predicted by random allocation of workers to establishments.”39 Hellerstein and Neumark write that available empirical evidence prior to their study suggests the presence of systematic job segregation along the lines of sex, race, and ethnicity. One study finds that “job cell segregation by race accounts for about half of the black-white wage gap.”40 This finding suggests there is ample room for passive discrimination to operate, though certainly specific evidence would be necessary to conclude that it does in fact operate.

Using high-quality, restricted-use, matched employer-employee data from the 1990 US Census, Hellerstein and Neumark write that “[o]ur results point to workplace segregation by education and race. . . . We find, however, that education plays very little role in generating workplace segregation by race.”41 This result suggests that at least one of Epstein’s “key differences” likely has little to do with racial job segregation, whatever association it has with racial wage gaps. Hellerstein and Neumark’s is of course just one study, and to our knowledge, an empirical study of the kind of segregation to be expected from passive discrimination has not been done. Nevertheless, we do not believe the possibility of workplace segregation as induced by passive discrimination is far-fetched.

As a final source of empirical evidence, consider a recent audit study conducted by Marianne Bertrand and Sendhil Mullainathan,42 who sent fake resumes to real employers who advertised jobs for new employees. Pairs of resumes were identical except for one feature: the authors “experimentally manipulate perception of race via the name of the fictitious job applicant . . . [by] randomly assign[ing] very White-sounding names . . . to half the resumes and very African-American-sounding names . . . to the other half.”43 They found “large racial differences in callback rates.”44 They note white-named applicants need to send about ten resumes per callback, whereas African-American-named applicants must send about fifteen resumes. They conclude, “This 50-percent gap in callback is statistically significant. A White name yields as many more callbacks as an additional eight years of experience on a resume. Since applicants’ names are randomly assigned, this gap can only be attributed to the name manipulation.”45 In sum, we believe the empirical evidence on racial gaps in the labor market points to a less sanguine conclusion than Epstein’s. We agree with him that explanations for racial “disparity are likely to prove complex,” but we do not believe that the available evidence supports his blanket claim that “invidious discrimination on racial grounds does not look to be one of them.”


In conclusion, while we are grateful to Professor Epstein for taking the time to comment on our Article, we will have to agree to disagree about a number of the premises upon which many of his arguments are based. As for Epstein’s larger complaints about the existence and effects of Title VII, for brevity’s sake, we defer to the wider employment discrimination literature which we feel adequately addresses these claims.


Copyright © 2010 University of Chicago Law Review.

Jonah Gelbach is Associate Professor of Economics at the University of Arizona.
Jonathan Klick is a Professor of Law at the University of Pennsylvania Law School.
Lesley Wexler is an Assistant Professor at Florida State University School of Law.

  1. See Jonah Gelbach, Jonathan Klick, and Lesley Wexler, Passive Discrimination: When Does It Make Sense to Pay Too Little?, 76 U Chi L Rev 797, 799, 802 (2009).
  2. Richard A. Epstein, Protect Us, Lord, from Title VII: A Response to Gelbach, Klick, and Wexler, U Chi L Rev Legal Workshop (June 22, 2009), online at (visited Dec 26, 2009) (footnote omitted).
  3. Of course, choice has potential second-order downsides such as pooling problems and bad choice, as we mention in our footnotes. These may be more significant than the risk of segregation, but our point in the original Article was merely that choice solves the segregation and valuation problem.
  4. Gelbach, Klick, and Wexler, 76 U Chi L Rev at 853–54 (cited in note 1) (emphasis added).
  5. Id at 853–56.
  6. Epstein, Protect Us, Lord, from Title VII (cited in note 3).
  7. Id.
  8. Id.
  9. Of course, Epstein overlooks the possibility that an employer may want to sort among different groups of minority workers. We raise this option when we discuss the use of language policies to screen in subservient workers and screen out more litigious minorities. See Gelbach, Klick, and Wexler, 76 U Chi L Rev at 821 (cited in note 1).
  10. Id at 801.
  11. Such datasets would need to contain a great deal of detail about employers’ compensation packages.
  12. Emily Schmall, The Cult of Chick-fil-A, (July 23, 2007), online at (visited Dec 26, 2009) (noting at least twelve charges of employment discrimination have been filed against Chick-fil-A since 1988, including a suit that was settled in 2000 after a Muslim manager was fired after refusing to participate in a group prayer at a company training program in 2000). We recognize the mere existence of suits is not per se evidence of discrimination, but it is highly suggestive.
  13. Chick-fil-A appears to care a great deal about screening hires and operators, including for many, a yearlong vetting process that includes dozens of interviews. Ty Yokum, the training manager . . . , sat through 7 interviews and didn’t get the job. He reapplied in 1991 and was subjected to another 17 interviews—the final one lasted five hours—and was hired. . . . Chick-fil-A’s general counsel[
  14. That best policy would be to “[j
  15. Id.
  16. Id.
  17. Id.
  18. Epstein, Protect Us, Lord, From Title VII (cited in note 3).
  19. The full process would go as follows. Given free entry into production, either via entry of new firms or expansion by existing firms, an industry’s output price will be competed down to minimum long-run average cost. However, in a standard model, a firm cannot both pay a premium to attract favored workers and produce at minimum long-run average cost, since attracting only favored workers requires paying them a premium. To get marginal revenue product to equal this above-market wage, a discriminating employer must employ a number of workers that induces a greater marginal product of labor than the number of workers employed by a nondiscriminating employer. (This is true given that the employer is a price taker on her output market.) Since nondiscriminating employers are cost minimizers, and since discriminating firms behave differently, a discriminating employer cannot be using the long-run cost-minimizing input mix. Thus, discriminating firms will produce at a cost above minimum long-run average cost. Given free entry, this means a cost above the long-run equilibrium price. As a result, such firms will go broke and exit the market, leaving only nondiscriminating firms.
  20. The formal condition for a monopsonist employer to maximize profits in choosing its quantity of labor and wage requires that marginal revenue product equal marginal factor cost. In a competitive market, marginal factor cost is the same as the wage. For a monopsonist, though, the profit-maximizing quantity of labor is always low enough that the wage will be less than marginal revenue product, leaving room for the employer to pay a premium to favored workers.
  21. Epstein, Protect Us, Lord, from Title VII (cited in note 3).
  22. Social Security Administration Historian’s Office, Research Note #1: Origins of the Three-Legged Stool Metaphor for Social Security (May 1996), online at (visited Dec 26, 2009).

  23. 435 US 702 (1978).
  24. Epstein’s focus on perfect monopsony power also seems unduly restrictive. Employers needn’t be the only game in town to be able to have equilibrium market power. The key question in nonperfectly competitive markets concerns the ease of entry. When entry is difficult, even firms facing some competition on either their input or output markets can discriminate in long-run equilibrium. On the other hand, in the monopolistic competition case with free entry, firms have pricing power in the short run, but not the long run. An employer that sought to discriminate against some workers in a monopolistically competitive industry with free entry and homogeneous production technology would ultimately go broke, as in a perfectly competitive market. This example shows that the necessity of antidiscrimination laws hinges on the specifics of industry conditions.
  25. Epstein, Protect Us, Lord, from Title VII (cited in note 3).
  26. Id.
  27. Nothing important in this discussion is altered by the fact that in our model, labor market equilibrium allows (and sometimes requires) cash-and-fringe compensation to replace cash-only compensation. When fringe benefits are an available element of compensation packages, the marginal cost to the firm of total compensation replaces the wage in the marginal revenue product-equals-wage condition. The only other modification is that in equilibrium, competitive firms cannot pay workers too much fringe: the compensation package must allow workers to maximize their welfare subject to the total compensation the firm pays.
  28. Epstein, Protect Us, Lord, from Title VII (cited in note 3).
  29. Id.
  30. Id (footnote omitted).
  31. Epstein focuses on the difference across sex in conditional black-white wage gaps. (A conditional gap is one computed using statistical methods to control for differences in characteristics besides the one used to define the groups whose gap is measured.) The study he cites uses data only on college graduates. See Epstein, Protect Us, Lord, from Title VII (cited in note 3). Thus, its design prevents it from telling us anything empirical about workers with less educational attainment. Since blacks and whites have very different educational attainment on average, it concerns disproportionately fewer African Americans than whites. As a result, Epstein’s favored study cannot be generally dispositive on this larger issue.
  32. Derek A. Neal and William R. Johnson, The Role of Premarket Factors in Black-White Wage Differences, 104 J Polit Econ 869 (1996).
  33. These facts can be seen by reference to Neal and Johnson’s Table 1, on page 875 of their article. In column (4) of this table, Neal and Johnson control for neither education nor their measure of test scores (the Armed Forces Qualification Test, or AFQT), and they find that black women in their sample earn 18.5 log points less than white women in their sample do; this estimate is highly statistically significant against a null hypothesis of zero difference. In column (6) of this table, Neal and Johnson add as an additional regressor a measure of each person’s AFQT score, and they find that black women earn 3.5 log points more than white women; this estimate is insignificantly different from zero, however, which motivates our phrasing in the main text concerning the black-white wage gap among women. Analogous results among men in Neal and Johnson’s sample appear in columns (1) and (3) of their Table 1. These estimates show that blacks in the sample earn an estimated 24.4 log points less than do whites in the sample when AFQT is not included as a covariate, but only 7.2 log points less when AFQT is included (both estimates are statistically significant). This amounts to a 100%×(1-7.2/24.4) = 70.5% reduction in the baseline estimate of 24.4 log points.
  34. Kevin Lang and Michael Manove, Education and Labor-Market Discrimination (unpublished manuscript, Boston University, Feb 2008), online at (visited Dec 26, 2009).
  35. Id at *29.
  36. Joseph G. Altonji and Rebecca M. Blank, Race and Gender in the Labor Market, in Orley Ashenfelter and David Card, eds, 3 Handbook of Labor Economics 3143 (1999).
  37. Id at 3164.
  38. Judith K. Hellerstein and David Neumark, Workplace Segregation in the United States: Race, Ethnicity, and Skill, 90 Rev Econ & Stat 459 (2008).
  39. Id at 459.
  40. Id.
  41. Id at 461.
  42. Marianne Bertrand and Sendhil Mullainathan, Are Emily and Greg More Employable Than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination, 94 Am Econ Rev 991 (2004)
  43. Id at 992.
  44. Id.
  45. Note that some (or all) of this effect is not necessarily the result of invidious discrimination: employers may have been exploiting correlations between a person’s name and unobservable characteristics that influence a worker’s marginal product of labor. On this point (and as a general criticism of audit studies), see James J. Heckman, Detecting Discrimination, 12 J Econ Persp 101, 107–11 (Spring 1998). However, the study at least suggests the possibility that employment discrimination endures in modern labor markets. It also demonstrates once again the difficulty of distinguishing between intentional and unintentional passive discrimination.
  46. Epstein, Protect Us, Lord, from Title VII (cited in note 3).

Post a Comment (all fields are required)

You must be logged in to post a comment.