Chapter 1: "Untitled" Discrimination

In a comedy sketch broadcast many years ago on Saturday Night Live, Eddie Murphy (a black comedian) disguised himself to look like a white person only to discover that white consumers are treated radically better than black consumers. At one point in the sketch, the seemingly white Murphy attempts to pay for a magazine at a store only to be told by the store keeper that there is no need to pay so long as black customers are not around. The skit confirmed a deep suspicion of people of color -- that they could uncover pervasive discrimination if only they were privy to knowledge of how whites are treated.

A PrimeTime Live segment broadcast several years ago had a similar punch line. The news magazine sent a white male and black male separately to shop into a variety of retail stores along with a hidden camera to record how each was treated. The video tape depicted several different types of race discrimination. A car salesman offered the black male a higher price; in an empty shoe store a clerk forced the black "tester" to wait for 15 minutes before assisting him; and in a record store a clerk conspicuously trailed the black tester as he browsed, while the white tester was allowed to shop without scrutiny.

In the Saturday Night Live sketch, the whites knew about the discrimination and blacks were kept in the dark. But I believe the PrimeTime Live segment suggests that something closer to the opposite is true. Several white colleagues and students who have seen the segment have remarked how it opened their eyes to the reality of quotidian discrimination in retail markets.(1) People of color have better insights into what normal service is and hence are better attuned to the possibility of race discrimination. But knowledge about how other similarly situated people are treated is a crucial barrier to learning whether (and the extent to which) discrimination exists. Consumers usually know how they are treated, but often lack information on how other consumers are treated. Equality of treatment can thus be what economists call a "credence" attribute--that is, a characteristic of consumption that the consumer does not learn about either at the time of sale or through consumption of the product or service itself.(2) Just as a manufacturer's claim that its lightbulbs have an expected life of 2,000 hours, a car dealer's explicit or implicit claim that it does not discriminate on the basis of race or gender may be difficult for consumers to verify (through their individual experience).

Of course, there are many types of discrimination about which consumers would readily learn if they systematically occurred. For example, if a McDonald's franchise set a higher hamburger price for Hispanics than Anglos, it is likely that customers waiting in line would observe the disparate treatment.(3) We should not expect hamburger price discrimination to persist in a competitive market place.

Indeed, there seems to be a widespread, implicit belief (at least among Caucasian males) that race and gender discrimination is not a serious problem in retail markets. The civil rights laws of the nineteen sixties focused on only a handful of non-retail markets--chiefly concerning employment, housing, and public accommodation services.(4) Indeed, the most gaping hole in our civil rights law concerns retail gender discrimination. No federal law prohibits gender discrimination in the sale of goods or services. A seller could flatly refuse to deal with a potential buyer of a car or a paperclip because of her gender. And while the civil rights laws of the eighteen sixties prohibited race discrimination in contracting, the civil rights laws of the nineteen sixties only prohibited sex discrimination in a narrow range of "titled" markets. The thousands of other markets which make up our economy are completely unregulated with regard to gender (as well as religion and national origin) discrimination and less regulated with regard to race.(5) And only a handful of states and cities (chief among them California) make up for this failing by prohibiting gender discrimination in contracting generally.

The non-regulation of retail discrimination seems to be premised on a vague coterie of beliefs that (1) retail discrimination does not exist because retailers have no motive to discriminate; (2) retail discrimination does not exist because competition forces retailers not to discriminate; and (3) any retail discrimination that does occur does not have serious consequences because of effective counterstrategies by potential victims. It is also argued that any discrimination in the sale of goods or services is less important than the potential effects of discrimination in the markets for employment and housing. But without denying the primacy of employment, the current regulatory regime leaves approximately 66% of dollars we spend -- and 35% of the dollars we earn -- unregulated (with respect to gender discrimination) or less regulated (with respect to race discrimination).(6)

In this book, I hope to contest the idea that race and gender discrimination in the retail sale of goods is non-existent or unimportant. My thesis is that race and gender discrimination is not a thing of the past, nor is it limited to the narrow set of "titled" markets regulated by civil rights legislation of the 1960s (Title VII, Title II, etc.). The book's primary contribution is empirical, but let me begin with a few theoretical reasons why we should take the possibility of retail discrimination seriously.

Retailers May Have a Motive To Discriminate. The argument that discrimination in the sale of goods and services does not exist because retailers lack any disparate treatment motive is itself premised on the twin ideas that discriminating against economically marginal groups would not be profitable and that racial animus would not manifest itself in discrete retail transactions. The latter idea is that while animus might cause race discrimination in the more relational employment, apartment rental, and restaurant settings (which are covered), uncovered retail transactions are sufficiently discrete that seller and/or customer prejudice would not induce disparate treatment. There are, however, several problems with this theory. First, as pointed out by Ian MacNeil, many contractual arrangements are not as discrete as initially appear.(7) Barbers may have much more tactile and repeated contact with their customers than one-time sellers of a house, but the law much more vigorously regulates the latter transaction. Second, the thought that prejudice is less likely to to be acted upon in discrete transactions is premised on a narrow theory of what might be called "associational" animus--that is, that bigots don't like associating with particular groups. But, as I will discuss in Chapter 3, there are other types of animus that might persist even in discrete markets. For example, if sellers enjoy extracting an extra dollar of profit from people of color more than from Caucasians, we might expect to see disparate racial treatment in pricing or quality of service. Finally, appreciating the pervasive discretion given to employees as agents opens up the possibility that even profit-maximizing principals will by necessity countenance some disparate treatment by their subordinates.

Discriminating retailers may also be actuated by profit. Sellers may have a profit-maximizing incentive to price discriminate against minorities and women--even if sellers believe that members of these groups are on average poorer. It has long been known that "statistical discrimination" might cause rational, profit-maximizing sellers to charge more to groups that on average cause sellers to incur higher costs. Thus, as a theoretical matter, the drivers of taxis might discriminate against African-American men if the drivers perceive a higher chance of being robbed by such passengers.(8) But, more provocatively, focusing on the example of new car sales, I will argue that profit-maximizing sellers may engage in "revenue-based" race and gender disparate treatment. Dealerships may discriminate not because they expect higher costs but because they expect to be able to extract higher revenues. This is a surprising possibility because as an empirical matter people of color have a substantially lower ability to pay for new cars. But profit maximizing sellers care far more about the variability in willingness to pay than in the mean willingness. The presence of a few minority members who are willing to pay a large markup can make it rational for the dealership to offer higher prices to all members of the group--even if group members are on average poorer.

It is correct and useful to ask whether sellers would plausibly be motivated to engage in a particular type of discrimination. But treating this issue seriously opens up a variety of dimensions where discrimination in the retail sale of goods and services could be a plausible seller strategy for either profit or non-profit based reasons..

Competition May Not Drive Out Retail Discrimination. Nobel-prize winning economist Gary Becker emphasized how competition could provide a great antidote for the disease of discrimination.(9) Nondiscriminating sellers could earn higher profits by picking up the sales of those minorities and/or women excluded from equal access to the discriminating sellers. One problem with this theory is that it focuses on the ability of competition to drive out discriminating sellers, but competition may not be as effective at driving out the effects of discriminating customers. If fixed costs of production limit the number of firms selling and if a substantial number of, say, Caucasian customers prefer dealing with a firm that discriminates against (by excluding or offering inferior service to) people of color, then firms may decide that it is more profitable to exclude minorities than to lose the patronage of Caucasians.

However, as an empirical matter, my guess is that most firms would not find overt race or gender discrimination to be a profit maximizing strategy. While Lester Maddox may have increased his sales by excluding African Americans,(10) in most markets a "whites only" or "males only" policy or overtly charging higher prices to particular demographic groups would lead to a general negative consumer reaction--by both minorities and progressive white consumers.

A more important limitation on competition is consumer information. In order for discrimination to cause the competitive shift of consumers toward nondiscriminatory sellers, it is important for consumers to know which sellers are discriminating and which are not. There is thus an important informational prerequisite for competition to have the predicted Beckerian effect. But as described above, there are many aspects of treatment where consumers may not be able to compare how sellers treat similarly situated counterparts.

Retail discrimination is most likely to persist where consumers do not learn the benchmark treatment of fellow consumers. Thus, while there is little opportunity for a single fastfood franchise to charge different prices for hamburgers, it should be possible for a dealership to charge different prices for cars. Since bargained prices diverge from the list price, it is very difficult for a consumer to know whether she has received a non-discriminatory price. And it will be more difficult for a nondiscriminatory seller to credibly market itself on the basis that the race or gender of customers do not influence its bargaining strategy.(11)

Markets in which price or other terms of trade are individually bargained provide much greater opportunities for race or gender discrimination than markets with homogeneous products attributes and posted prices. However, even retailers that sell standardized products at posted prices might discriminate on the basis of race or gender with regard to discretionary aspects of service. Anyone watching the Prime Time segment could vividly see that record and department stores could substantially increase the "transaction costs" of minority customers. This is not just an issue of whether the retailer provides "service with a smile" but, as in the PrimeTime Live testing, whether the retailers make minority customers wait substantially longer periods before being served (or whether the minority customers are conspicuously shadowed to scrutinize whether they are shoplifting).

Retailers may also discriminate in their willingness to accommodate private and somewhat idiosyncratic consumer requests. For example, Jane Connor is currently testing retailers in Binghamton, New York to see whether there are racial differences in their willingness to accede to a request to use a restroom or a request to return a sweater without a receipt. Economists (and others) tend to ignore or downplay the harms of such discrimination. But non-trivial injury may be visited on people of color in terms of both higher transactions costs and taking more precaution to comply strictly with retailer policies. One audit study showed that African Americans in Washington, D.C. had to wait 27% longer to hail a cab.(12) If this seems a minor inconvenience, white readers should try to imagine what their life would be like if every (or even just many) transactions took 27% longer. Even a single incident can impose real psychological costs. Consider, for example, Patricia Williams' story of being denied entrance to an open Benneton store by a gum-chewing, buzzer-wielding young white store clerk?(13)

While the term "public accommodations" refers to particular types of markets covered by antidiscrimination laws, Connor's research suggests that retailers' willingness to make "private accommodations"may be an important location where disparate racial or gender treatment may occur. And because such accommodations are either made privately or arise in non-standard circumstances, it is unlikely that competition will be an effective palliative. By the one-off nature of these transactions, it is difficult to infer whether a particular retailer action is motivated by the requesting consumer's race or gender, and consequently it will be difficult for nondiscriminating retailers to market themselves as such to people of color and/or women.

The point here is that imperfect consumer information about whether a particular retailer discriminates can impede the competitive responses necessary to drive out discrimination. Ironically, one of the world's leading econometricians, James J. Heckman, has recently criticized audit testing itself because of an analogous problem. Heckman writes:

[T]he evidence acquired from [audit pair tests] is less compelling than is often assumed. Inferences from such studies are quite fragile to alternative assumptions about unobservable variables . . . .(14)



In essence, Heckman is arguing that because social scientists will not be able to observe all the factors that might plausibly motivate a decisionmaker, it is unwarranted to infer that disparate treatment was due to testers race or gender as opposed to some other unobservable factor. If it is difficult for social scientists to discover in purposive, controlled and systematic testing whether a seller is discriminating on the basis of race or gender, however, it will be all the harder for a consumer to acquire such information. I disagree with Heckman that the results of such studies are "quite fragile" to alternative assumptions (and Heckman provides no citations to support this claim). Particularly in retail settings, it is possible for social scientists to control for the most plausible attributes that could credibly affect seller behavior.(15) In the next chapter, I will discuss the details of my audit test at new car dealerships in which I controlled for a host of different testing attributes--including not just speech, clothing, and physical attractiveness of the testers, but also the type of car that the testers used to approach the dealership. Of course, Heckman is right that I can't be sure the testers blinked their eyes at the same rate, but I doubt that results are fragile to controlling for other factors. However, there is an important point of agreement in Heckman's and my analysis, and that is to see that imperfect information about whether discrimination exists can insulate the practice from either public or private responses.

Victims Self-Help May Not Eradicate the Injury. Even if competition is not strong enough to drive discriminatory treatment from the market, the ability of potential victims to purposively select the retailer from which they will purchase may importantly mitigate the harm from discrimination that people of color bear in equilibrium. Again, Heckman has championed this point:

[Audit evidence of disparate racial treatment] is entirely consistent with little or no market discrimination at the margin. Purposive sorting within markets eliminates the worst forms of discrimination. There may be evil lurking in the hearts of firms that is never manifest in consummated market transactions.(16)



Heckman is clearly correct as a theoretical matter that purposive sorting can mitigate the effects of disparate treatment by individual retailers. But the above quoted assertions inappropriately transform--without the benefit of any data--theoretical predictions into descriptions of empirical fact. There is simply no evidence on the degree to which purposive sorting mitigates the harms of disparate treatment, and, indeed, I will provide evidence in Chapter 4 that disparate racial treatment does persist in "consummated market transactions" when I examine actual car sales.

Moreover, there is a deep tension between Heckman's previous assertions that social scientists are unable--even with the aid of controlled audit tests--to identify which retailers are discriminating and his assertion that consumers can purposively shift to retailers they know to be non-discriminating. The same unobservable factors that make identifying discrimination difficult for social scientists will impede consumer self-help.

At the beginning of this chapter, I suggested that--counter to the Eddie Murphy skit--blacks who have better information about "normal" treatment are likely to have a better inkling than whites about the degree of disparate treatment occurring in particular retail markets. It is instructive to note that in a 1997 Gallup study, 30% of blacks surveyed said that they had experienced discrimination while "shopping" in the last 30 days.(17) By way of contrast, the figure for experiences of discrimination while "dining out" (including bars, theaters and other entertainment) was only 21 percent. This suggests that discrimination in the purer "public accommodation" markets may be less severe than the "untitled" retail markets generally. But having a better inkling about the existence of discrimination is different than knowing (a) whether a particular instance of treatment is inferior to that of other customers and (b) whether that disparate treatment is caused by your race. Retailers may still be able to maintain "plausible deniability" about whether a failure to accommodate a particular request is an aberration or a manifestation of systematic discrimination. With regard to idiosyncratic or infrequent accommodations, it will be hard even for minority consumers to know the true basis of their treatment and the true benchmark. Retailers can compete on price, but it will be much harder for them to compete on "service with a smile" and even harder to compete on implicit promises of non-discrimination. Just as HMOs have difficulty competing on whether they will aggressively treat cancer,(18) it is difficult for retailers to compete on whether you will be treated well if you lose a receipt or if your toddler needs to use the employee bathroom.

But my objective here is not to deny that either profit-motived sellers or purposively sorting consumers can be powerful forces against private discrimination. Instead, I hope to show that taking theories of motive and knowledge seriously allows us better to predict where discrimination can persist as well as situations where it cannot. The use of the term "Pervasive Prejudice?" in the title is not meant to imply that race and gender discrimination pervades all aspects of market behavior, but instead is meant to indicate that discrimination may occur in a wide range of retail markets. Notwithstanding competition, race and gender discrimination can persist not just in the "titled" markets, but in any of the "untitled" retail markets where minority consumers have imperfect information about how their white counterparts are treated. The question mark in the title should also be emphasized. To demonstrate empirically that race or gender discrimination occurs in a wide range of markets would require amassing evidence from hundreds of different product or service markets. This book does not accomplish this task, but instead provides a more intensive investigation of a smaller set of markets. Peter Siegelman and John Yinger have written separately the two most comprehensive review articles on retail discrimination.(19) My hope is that providing detailed evidence of race and gender discrimination in specific geographic and product markets--particularly with regard to automobile sales in the Chicago area--will make the pervasiveness claim a more plausible policy concern.

The subtitle's reference to "non-traditional evidence" also deserves comment. The empirical tests reported in this book are non-traditional in both method and subject matter. While civil rights testing has traditionally focused on employment, housing, and public accommodations, this book provides evidence of discrimination in a variety of "non-traditional" markets, including retail car sales, kidney transplantation, bail bonding, and FCC licensing. The four major empirical findings are:





The methodology of investigating these markets is also often non-traditional in introducing new econometric and game-theory procedures for analyzing civil rights issues. The three major methodological contributions of the book are:





Together, these last two points suggest that game-theory is an underutilized tool for civil rights practitioners that can help predict how different causal theories (animus-based, statistical discrimination, etc.) and different government remedies (affirmative action bidding credits) would affect private behavior.

The book is divided into three parts--concerning disparate treatment, disparate impact, and affirmative action, respectively. The first part provides an intensive exploration of race and gender disparate treatment in new car sales. Chapter 2 describes the basic audit finding of discrimination. Chapter 3 explores alternative causal explanations for the disparate treatment. Chapter 4 contrasts the audit results with subsequent evidence of discrimination in consummated transactions, and Chapter 5 discusses legal implications of the analysis.

Part II provides two contrasting tests of disparate racial impact. Chapter 6 provides traditional evidence that the government kidney transplantation "point system" has had an unjustified disparate impact on African Americans. To accomplish this, I show first that the point system disproportionately excluded African Americans as transplantation recipients and then separately demonstrate that the point system was not justified in terms of enhancing the expected transplant success rate. In contrast, Chapter 7 uses the "outcome" test to show that the criteria used by New Haven judges in setting bail bonds has had an unjustified disparate impact on black male defendants. The market evidence that black male defendants have a lower probability of flight suggests that the criteria for setting the amount of bail bonds disproportionately overdeters these defendants relative to white males. Finally, Part III uses game theory to inform its tests of how affirmative action bidding credits affected the ultimate sale price in recent FCC narrowband spectrum auctions. The concluding chapter focuses on information and motives as a fruitful way of grounding future empirical scholarship and policymaking in our continuing struggle to identify and eliminate discrimination from our economy.

Five of the book's nine chapters are based on research that originally appeared in article form. But this book is not merely a concatenation of my previously published empirical articles on race and gender discrimination. Rather, where ever possible I have tried to update the original empiricism with additional testing (as with Chapter 4's new analysis of dealer discrimination in consummated transactions), and to provide broader theoretical implications for the methodology (as with Chapter 9's extended discussion of outcome tests of unjustified disparate impact).

In three of the chapters (Chapter 5 concerning car discrimination, Chapter 6 concerning kidney discrimination and Chapter 8 concerning FCC auctions), I explicitly respond to critics of the original articles. In these chapters, I also discuss how subsequent regulation (for example, of kidney transplantation and of radio spectrum auctions) or technological innovation (such as the emergence of the Internet in car sales) has changed the opportunities for discrimination. At times, I have chosen to retain some of the text of the original articles - both so that readers can get a sense of the original arguments that provoked the subsequent criticism and so that readers will understand the technological and regulatory constraints that potential discriminators faced at the time of their decision-making. But in what follows, I will endeavor to make clear what information was available at the time of the decisions being studied and how this information may have subsequently changed.

Chapters 2 and 3 -- on discrimination in new car sales -- rely on Fair Driving: Gender and Race Discrimination in Retail Car Negotiations, 104 Harv. L. Rev. 817 (1991); Race and Gender Discrimination in Negotiation For the Purchase of a New Car, 84 Am. Econ. Rev. 304 (1995) (with Peter Siegelman); and Further Evidence of Discrimination in New Car Negotiations and Estimates of Its Cause, 94 Mich. L. Rev. 109 (1995). Chapter 6 -- on kidney transplantation -- relies on Racial Equity in Renal Transplantation: The Disparate Impact of HLA-Based Allocation, 270 J. Am. Med. Assoc. 1352 (1993) (with Robert Gaston, Laura Dooley, and Arnold Diethelm); Response to letters-to-the-editors, 271 J. Am. Med. Assoc. 269 (1994); Unequal Racial Access to Kidney Transplantation, 46 Vand. L. Rev. 805 (1993) (with Laura Dooley and Robert Gaston); and HLA Matching in Renal Transplantation, 332 New Eng. J. Med. 752 (1995) (with Robert Gaston and Mark Deierhoi). Chapter 7 -- on bail bond setting -- relies on A Market Test for Race Discrimination in Bail Setting, 46 Stan. L. Rev. 987 (1994) (with Joel Waldfogel). And Chapter 8 -- on FCC spectrum auctions -- relies on Pursuing Deficit Reduction Through Diversity: How Affirmative Action at the FCC Increased Auction Competition, 48 Stan. L. Rev. 761 (1996) (with Peter Cramton) and Aid Diversity, and the Treasury, N.Y. Times, at F13 (May 21, 1995) (with Peter Cramton).

1. While visually compelling, the PrimeTime Live segment suffered from the important methodological flaw that it failed to represent whether or not the images of discrimination depicted were representative of the overall testing. Particularly, the producers failed to indicate in what proportion of cases the African-American tester was treated worse--as well as what proportion of the time the caucasian tester was treated worse. This failing is particularly egregious because such a disclosure would have taken only a few seconds of air time and because the producers were on explicit notice that such a representation was necessary. The PrimeTime Live segment was inspired by the news coverage surrounding my car discrimination work, discussed in Chapter 2. The producers of the segments telephoned me and asked for my input before conducting the tests but ignored my suggestions about "representativeness," both with regard to the initial race testings and also with regard to subsequent tests concerning gender and age discrimination.

2. Credence attributes of products are usually distinguished from "search attributes" (which consumers learn about before they purchase the product) and "experience attributes" (which consumers learn about as they consume the product). See, e.g., Richard Craswell, Interpreting Deceptive Advertising, 65 B.U. L. Rev. 657 (1985).

3. The posted pricing facilitates comparison. Minorities would immediately know if they were asked to pay more than the posted price. And would probably be able to observe if whites ahead of them in line were asked to pay less than the posted pirce.

Some French restaurants, however, charge higher prices on their menus translated into English than on their French counterparts. So long as the English tourists don't ask for a French menu as well, then they are likely to be unaware of the price discrimination.

4. The last category, public accommodations, is most elastic, see infra Chapter 2, but traditionally has included only producers who were regulated as "common carriers"--such as hotel, restaurants and means of public transportation.

5. Sections 1981 and 1982 prohibits only "intentional discrimination" on account of race, while Title VII and its counterparts also regulate conduct which has an unjustified disparate racial impact.

6. These admittedly heuristic percentages were calculated by crudely dividing the national income and expenditure accounts into covered and uncovered categories. For example with regard to expenditures, I assumed that "meals and beverages, purchased," "barbarshops, beauty parlors, baths, health clubs," "housing" (by far the largest), and "admission to spectator amusements" (including movie theaters, etc.) were covered. In 1985, 877 billion out of 2.6 trillion in personal consumption expenditures (34%) fell into covered categories. Statistical Abstract of the United States 1987 (Table 710). The income percentage was derived analogously by dividing personal income into covered categories (chiefly "wage, salary and other labor income") and uncovered categories (all other income). Id at Table 711.

7. See Ian MacNeil, The New Social Contract 10 (1980).

8. I emphasize the theoretical possibilities of cost-based statistical discrimination on the part of cab drivers should not imply that observed discrimination is consonant with rational inference. For example, there are numerous anecdotes of discrimination (particularly, refusals to pick up) against older African-American men and women dressed in suits and giving other indicia of low crime potential. See Cornel West, Race Matters xiv-xv (1993); see also Stanley E. Ridley et al., Taxi Service in the District of Columbia: Is it Influenced by Patrons' Race and Destination? (1989) (a report available from the Washington Lawyers' Committee for Civil Rights Under the Law). There are other potential statistical rationales for drivers' disparate treatment, including the "revenue-based" possibility that drivers believe they "will have a harder time finding a return fare from predominantly black neighborhoods that are the likely destination of black passengers." Peter Siegelman, Race Discrimination in "Everyday" Commercial Transactions: What Do We Know, What Do We Need to Know, and How Can We Find Out, (1998). An initial pilot study completed by Suzanne Perry and myself also suggests that black taxi patrons may leave substantially lower tips than whites, controlling for a host of other variables (including destination and length of trip). See also Farrell Bloch, Antidiscrimination Law and Minority Employment: Recruitment Practices and Regulatory Constraints 30 (1994).

9. See G. BECKER, THE ECONOMICS OF DISCRIMINATION 14-15, 39-54 (1957).

10. Lester Maddox gained national publicity shortly after passage of the 1964 Civil Rights Act when he distributed ax handles to supporters in order to prevent blacks from patronizing his Atlanta restaurant, the Pickrick. Ex-Governor Lester Maddox of Georgia has Heart Surgery, REUTERS, Apr. 1, 1991; Maddox to Campaign in Carolinas Against Dukakis, UPI, Oct. 17, 1988.

11. An important exception is the example of no-haggle dealerships, which will be discussed in Chapter 5.

12. Ridley et al., supra note 8.

13. See Patricia J. Williams, The Alchemy of Race and Rights 44-51 (1991). In an unpublished manuscript Devon Carbado lists 26 "privileges" that men and whites "enjoy" as they walk through their lives. Devon W. Carbado, Straight Out of the Closet (1999) (unpublished manuscript on file with author) .

14. James J. Heckman, Detecting Discrimination, 12 J. Econ. Perspectives 101 (1998).

15. Heckman's primary concern was with audit pair testing of employers in deciding whether to hire. Unobservable factors related to applicant productivity may be much more salient than in the retail context because of the more relational character of the employment contract and because the employee's promise of labor is much less standardized than a consumer's promise of money.

16. Heckman, supra note 14, at 103; see also Richard A. Epstein, Forbidden Grounds: The Case Against Employment Discrimination Laws 52 (1992).

17. See The Gallop Poll Social Audit on Black/White Relations in the United States (June 10, 1997) at 30-31 <www.gallup.com/poll/special/race.htm>. The poll was based on telephone interviews with a representative sample of 1269 blacks and 1680 whites who were interviewed in early 1997. The margin of error for a percentage estimate for blacks is approximately + 5 percentage points. See id. at 5-6.

18. See Russell Korobkin, The Efficiency of Managed Care "Patient Protection" Laws: Incomplete Contracts, Bounded Rationality, and Market Failure, 85 Cornell L. Rev. 1 (1999).

19. The most systematic survey of discrimination in consumer markets is Peter Siegelman, Race Discrimination in "Everyday" Commercial Transactions: What Do We Know, What Do We Need to Know, and How Can We Find Out, (1998). See also John Yinger, Evidence of Discrimination in Consumer Markets, 12 J. Econ Perspectives 23 (1998). Gender and race discrimination have been uncovered in a variety of markets. For example, several Washington, D.C. dry cleaners have discriminated against female customers by charging higher prices for women's blouses than for men's shirts. See Matlack, Experts on Call, 21 NAT'L J. 2549, 2549 (1989). Historically, blacks have been discriminated against in the sale of many goods and services. In 1959, black consumers and businesses associated with the NAACP were at times unable to buy such goods as milk, bread, groceries, gas, credit, fertilizer, seed, insecticides, and farm machinery. See American Friends Serv. Comm., Nat'l Council of Churches of Christ & S. Regional Council, Intimidation, Reprisal, and Violence in the South's Racial Crisis (1959), reprinted in Civil Rights--1959: Hearings Before the Subcomm. on Constitutional Rights of the Senate Comm. on the Judiciary, 86th Cong., 1st Sess. 1573 (1959); see also Perry v. Command Performance, 913 F.2d 99 (3d Cir. 1990) (involving the refusal by a beauty salon hairdresser to serve a black woman).