Essay

Discrimination by Customers

I.     The Conundrum

Customers discriminate, by race and gender, in every step of a market transaction. They discriminate in deciding with whom to do business. In one year-long experimental study, online advertisements for iPods showing the device in a black hand received 13% fewer responses and led to 17% fewer offers than the advertisements showing the device in a white hand.1 The same study also found that online buyers trust black sellers less than white sellers. Buyers were 17% less likely to include their name in e-mails to black sellers, 44% less likely to accept delivery by mail, and 56% more likely to express concern about making a long-distance payment.2

Customers also discriminate in how much they are willing to pay. An experimental study with baseball cards showed that, on average, a card in a black hand sold for 20% less than the same card in a white hand.3 Another study, based on the pictures and rental information of all New York City property owners seeking to rent their property on Airbnb, revealed that Airbnb guests are willing to pay non-black hosts approximately 12% more than black hosts for comparable properties,4 and exact a larger price penalty from black hosts for having a poor location.5 Discrimination by customers is well documented when it comes to tipping. A study of over 1000 taxicab trips in New Haven, Connecticut, found that customers tip black cab drivers approximately one-third less than white cab drivers, and that black drivers are 80% more likely to be stiffed.6 Studies of restaurant tipping show similar bias against female servers.7

Customers discriminate, as well, in how they evaluate those with whom they do business. The marketing company WordStream analyzed 300 responses from the company’s clients to a request to rate their level of satisfaction with client service representatives. Overall, clients undervalued female marketers by 21%.8 This pattern of discrimination is similar to what is observed when students rate their professors on teaching evaluations.9

Current law does little to address discrimination by customers, even though it disproportionately hurts women and persons of color. Federal and state anti-discrimination laws prohibit firms10 of a certain size from discriminating against employees and customers on the basis of race, gender, and other characteristics.11 Firms may not decide to hire, or serve, only whites. Yet the law does not prohibit discrimination by customers.12 White diners may frequent one café over another because they feel more comfortable with its white clientele. Clients may refuse to hire a female lawyer to represent them because they believe men are more competent. Customers may avoid doing business with a gay photographer because they are disgusted by homosexuality. They may move to another line at a car-rental counter to steer clear of a service representative wearing a Sikh turban because they assume all foreign-looking people are terrorists. They may tip white male waiters more than black female ones because they believe men are, and should be, the family breadwinners. Current law barely even applies to firms when they act as customers; while in the customer role, they, too, may bring their race and gender biases to bear on with whom they select to do business.13

It is a puzzle why we generally take for granted the right of customers to discriminate when they exercise their buying power. When society deems other market behaviors harmful enough, it customarily regulates both the supply and the demand side of the transaction. For example, both the sale of certain drugs and their possession are criminal.14 Prostitution laws make both selling and buying sex illegal.15 It is also illegal to both buy and sell body organs,16 as well as to offer and accept bribes.17

Why, then, do we consider discrimination unacceptable when engaged in by firms but allow it on the part of customers? There would be practical problems with a law aimed directly at customers, of course—problems in monitoring customer behavior, determining what behaviors are discriminatory, and proving a case of discrimination. As we later discuss in detail,18 these problems may be reason enough to steer away from a direct prohibition of discrimination by customers. Yet if discrimination by customers is a problem society wants to address, there are other approaches to consider. This Essay is the first academic work to address in detail whether there should be a public policy against customer discrimination and, if so, what legal approach makes the most sense.19

In Part II, we explore two significant explanations for why society might choose to ignore discrimination by customers. The first one is efficacy. Society may regulate firms (especially big firms) but not customers because it believes that the regulation of firms alone is sufficient to achieve society’s non-discrimination goals. The second explanation is a concern for personal autonomy and privacy. Society may regulate firms and not customers because it believes that limiting customer choice would infringe on the basic liberty interests of its citizens in ways not present in the regulation of firms. The concern with such potential infringement is not only that it potentially violates strong constitutional values, but also that legal coercion in the face of strong resistance to it may generate the kind of resentment that is counterproductive to the deeper improvements in attitudes that are necessary to reduce prejudice.

We conclude that while these two rationales might sensibly inform what legal approach to discrimination by customers is best, they do not justify ignoring this discrimination altogether. The efficacy rationale explains why regulating at the firm level may be more effective than regulating the behavior of individual customers, but not whether a regulatory scheme aimed exclusively at businesses is sufficient or optimal. The privacy rationale offers a more powerful justification for focusing on firms rather than on customers, but again does not justify the absence of any policy whatsoever for reducing discrimination by customers.

A policy against discrimination by customers would not mean necessarily treating all forms of discrimination the same. There are differences in the kinds of harms discrimination by customers impose, as well as in the relative benefits and burdens of strategies to reduce those harms. How society should approach discrimination by customers depends upon these differences. Although lines are hard to draw, in Part III we make an effort to distinguish between types of discrimination that are acceptable, perhaps even desirable; types that should be allowed, even if they have some potential for harm; and types that are harmful enough, and not justified by other factors, to warrant societal intervention.

Taking into account these differences, in Part IV, we sketch out a plausible policy approach. We conclude that a direct ban on discrimination by customers would be ineffective, and likely counterproductive. Instead, we advocate intervention directed at those upon whom the law already imposes statutory non-discrimination obligations—i.e. firms of a certain size. The proposal imposes a duty on these actors to refrain from facilitating harmful discrimination by their customers and, in some cases, to structure customer choices in order to diminish the effects of their discriminatory biases. This Part explains the difference this proposal might make, and explores the multiple strategies it might encourage firms to adopt.

We note that the matter of discrimination by customers is of growing importance as traditional boundaries within the economy are breaking down. Current law assumes a bifurcated market composed of firms who sell and individuals who buy; it regulates the former and not the latter. But firms in the new sharing economy—the Lending Clubs, Ubers, Airbnbs and Amazons of the world—operate not as sellers in the usual sense, but as coordinators of buyers and sellers. The sellers on these platforms are often individuals who seek to profit from their excess capacity, not corporate entities who produce goods and services to sell. Although expectations to complete a transaction in this sharing economy are also sometimes higher on sellers than on buyers,20 the actual buy–sell transaction online is between parties who are both customers of the platform, and thus beyond the scope of current anti-discrimination rules.21 This challenge to traditional transactional boundaries makes reconsideration of the law’s approach to customers especially timely.

II.     The Rationales for Regulating Only Firms

In this section, we explore the two most powerful reasons why society regulates discrimination by firms but not by customers: efficacy and privacy.

A.     Efficacy

 

1.     The Efficacy Rationale

 

One explanation for why we prohibit discrimination by firms and not by customers is that firms are easier to regulate. They are readily identifiable; they are already subject to many state obligations and regulations in exchange for the privilege of doing business in the state; and they have owners and managers who can be held accountable.

Firms are typically the cheapest cost avoiders. They participate in the market as a single player with control over how their employees are hired, promoted, and fired, and how their goods and services are produced and sold. Policies and procedures of the firm operate on a scale that individual consumer decisions do not. This feature not only gives firms more leverage than individual customers have; it also enables them to absorb the costs of ending discrimination more efficiently. To the extent that there are costs in refraining from discrimination—including the potential loss of revenues from bigoted customers—anti-discrimination mandates focused on firms tend to spread that loss throughout the industry in which the firms participate; without such spreading, firms choosing not to discriminate might be competitively disadvantaged.22

Another factor that makes firms the more efficient site of anti-discrimination obligations is that they are driven largely by profits, and thus already have an incentive to identify and eliminate potential inefficiencies within their organizations, including discrimination.23 In turn, eliminating their own discriminatory practices is likely to reduce opportunities for customers to discriminate. If businesses do not discriminate in their hiring and service practices, customers who do not want to sit at a lunch counter with blacks will have fewer places to dine out; airline passengers who prefer only female flight attendants will not have that choice if they want to fly; white customers who don’t want to shop alongside black customers will find their shopping choices narrowed. When customers have no choice but to deal with businesses with more diverse workforces, the resulting exposure to women and minorities may also tend to reduce customer biases toward these groups, thus altering the preferences that drive customer discrimination.24

 2.     Efficacy Limitations on Regulating Only Discrimination by Firms

If we had to choose who to regulate—firms or customers—the factors outlined above make a pretty good case for regulating firms. These factors do not explain, however, why the existing regulatory scheme focuses entirely on discrimination by firms, rather than on both firms and customers, nor why that regulatory scheme does not require firms to pay more attention to discrimination by its customers. Discrimination customer by customer may be less significant, on a decision-by-decision basis, than discrimination by firms, but it is harmful both in individual instances and in its cumulative effects. Moreover, just as we think that a benefit of eliminating discrimination by firms is that it lessens discrimination by customers, the converse is also true; if customers do not discriminate, neither, presumably, will businesses.

One could make the case that there is an even greater need for legal intervention to reduce customer discrimination than discrimination by firms. As noted above, firms tend to be profit-driven and thus already motivated to make rational, non-discriminatory decisions.25 Customers are more likely to be motivated by personal emotions, stereotypes, and preferences. Whether these motivations are conscious or implicit,26 customers typically act on them at little or no cost; in most markets, customers can find male lawyers or straight photographers or white cashiers to serve them. Even when there are costs,27 without having to account for profits in the same way firms do, customers may be more willing to pay for their discriminatory preferences than firms are.28 If this is the case, by focusing on discrimination by firms rather than by customers, current law concentrates on the slice of the market that is already strongly motivated to refrain from discrimination, and ignores players who do not.

The larger problem with a legal approach that targets only discrimination by firms is that it ignores the complex interaction between business practices and customer preferences.  Individual customer preferences and institutional practices do not operate separate and apart from each other. Instead, they interact in ways that create a feedback loop in which customer bias affects business decisions and business decisions reinforce customer bias. These phenomena will be examined in turn.

i.     How Customer Bias Affects Business Decisions

Customer bias influences decisions by businesses and the opportunities of their employees even when the businesses are not directly engaged in discriminatory behavior. Take the everyday example of customer service employees who face discrimination by customers. When customers steer away from store clerks wearing, say, Muslim garb, these workers will tend to generate fewer revenues for their employers, and thus may affect employer pay and promotion decisions. Prejudiced customers who have no choice but to deal with the Muslim workers will give bad evaluations,29 which reinforces these employment decisions. Thus, even if the firm (barred from discrimination) continues to interview and hire members of racial minorities, over time, rational clerks will tend to move themselves to the back office, learn to moderate the fact of their racial identity,30 or find a job (less well-paying) elsewhere. In this way, customer bias serves to lessen the market value of workers against whom customers are biased.

Another concrete example of customer bias affecting business decisions is the disparate pay apparent in male and female coaches in college sports. Colleges and universities often pay the coaches of their men’s basketball teams (mostly male, in part because of the preferences of the players) more than the coaches of female basketball teams (mostly female because, among other things, these jobs generally pay less).31 The jobs are arguably similar in skill, effort and responsibility—the factors that matter under the Equal Pay Act32—and a school’s women’s team is sometimes more successful than the men’s team.33 The usual justification for the pay differentials in coaching, which courts have accepted,34 is that men’s teams are revenue-producing and thereby involve working conditions with more stress and responsibility than women’s teams, which are not revenue-producing.35 This distinction explains the salaries of the most highly-paid coaches but it masks the fundamental dynamic at play: the reason men’s teams produce more revenue is that the public disproportionately prefers to watch men’s gamesi.e., customers discriminate.

One might argue that men’s sports display more talent, strength, athleticism, or speed and that these factors, not sex, explain any salary gap. But these characteristics are, themselves, gendered. As Deborah Brake writes:

[t]he popularity and revenue-producing potential of a sport . . . . is a product of countless social and institutional factors, including longstanding and continuing investments in facilities, personnel, programs, recruiting, marketing, and coaching . . . . These investments contribute to a certain image and status of a sport that greatly affect its marketability.36

To whatever extent the difference between male and female sports relates to factors other than gender, the fact remains that it is primarily because fans prefer men’s to women’s sports that men’s sports generate more responsibility, stress, and public relations work, and thus more pay for the coaches of men’s teams.37 The higher market rewards that flow to male athletes, and to male teams, not only disadvantages female athletes, but sends a message about the comparative value of male sports that reinforces these preferences.

Another example of how current law permits businesses to act on, and thus perpetuate, the stereotyped-based prejudices of their customers is the discretion allowed to businesses to impose gender-based dress and appearance standards on their employees.38 In most cases, employers are allowed to incorporate the sex-based expectations of their customers into how they require their employees to look. As long as the standards track community expectations and do not impose manifestly unequal burdens on male and female employees, they are not considered discrimination based on sex.39 Here, again, an employer is expected to refrain from acting on its own biases, but has considerable leeway to conform its practices to the biases of its customers.

Even firm practices that are understood to discriminate based on sex may be excused under Title VII’s Bona Fide Occupational Qualification (“BFOQ”) exception, under logic tied directly to customer preferences.40 For example, several cases have found sex to be a BFOQ in therapeutic settings in which customers expressed strong gender preferences.41 In one case, after nine female residents in a female nursing home signed an affidavit objecting “most strenuously” to male nurses and male nurse’s aides (although apparently not to male doctors), the nursing home adopted a policy of hiring only female nurses; this policy was upheld by the court under the BFOQ exception.42 In this way, the law treats customer biases as a given that firms may indulge when they can tie those biases to a business purpose.43

 ii.     How Business Practices Reinforce Customer Bias

Not only do customer biases influence business practices; in addition, business practices influence customer preferences. To an important extent, “[customers] do not come to the [market] with fully formed preferences”;44 these preferences emerge, based in part on the choices made available to them by firms. Thus, while firms theoretically have no business reason to discriminate except for the preferences of their customers, these preferences are often ones that the firms, themselves, help to create and nurture. The result is a complicated collusion in which discriminatory preferences perpetuate themselves without any clearly identifiable starting point.

This collusion is especially apparent in advertising. Advertising both reflects customer preferences and creates or reinforces those preferences. The most successful advertising is that which captures and builds on the emotional needs and desires of the customer and then associates specific products with the fulfillment of those needs. This process is fundamentally a social one. Customers, who may not even be aware of their preferences, are attracted to images of people with whom they seek to identify.45 Firms understand these associational preferences, and seek to exploit them. The Marlboro Man ads—sometimes described as “[t]he most powerful . . . brand image of the century”46—succeeded because potential customers wanted to be identified with the rugged, masculine values the cowboy image represents; for those to whom it appeals, smoking is not only acceptable, but also sexy and macho. Likewise, the clothing ads of many firms both exploit and construct female body images. An infamous example were the Abercrombie & Fitch ads depicting young, skinny girls in seductive poses, as well as tight t-shirts emblazoned with such slogans as “Who needs a brain when you have these?” and “Female students wanted for sexual research.”47 Ads that build on racist identity also tap into and reinforce racial stereotypes and prejudice. Consider an ad for “Wong Brothers Laundry Service: Two Wongs Can Make It White;”48 an Intel ad with six muscular black runners bowing in front of a white man dressed in business attire under the headline “maximize the power of your employees;”49 and a series of liquor ads depicting white people acting out black hip-hop stereotypes.50

While such ads may be the subject of public criticism or boycotts,51 they are not illegal. Wilson v. Southwest Airlines52 held that Southwest Airlines could not deny employment to male flight attendants, but it did not disapprove of the company’s advertising strategy, designed to distinguish it from other airlines by “project[ing] to the traveling public an airline personification of feminine youth and vitality.”53 Indeed, there was, and still is, no law under which the campaign could have been disapproved, and First Amendment considerations likely prevent the state from enacting one. Even in its new 2015 advertisements (toned down for the times), Southwest Airlines continues to be the LUV airline, catering to the positive customer response to its sexy imaging.54 Firms are well aware that they aren’t legally responsible for the prejudice their ads foster or discriminatory harms they cause.55

The influence of business practice on customer bias is especially apparent in the online economy. As noted above,56 online platforms such as Airbnb, Craigslist, Eatwith, Etsy, Rover and many other platforms do not initiate transactions as either sellers (who would be regulated) or buyers (who are not); rather, they facilitate the transactions of site users—sellers and buyers—by bringing them together. As facilitators rather than sellers in the usual sense, they can disclaim responsibility for the discrimination any of these users apply in their searches. Current law allows this.57

While online sharing platforms do not participate directly as sellers or buyers in a transaction, they amass data about their users, both sellers and buyers. This data is used to target site users with products and services in which their search and buying histories suggest potential interest.58 In directing their users to these products and services, these platforms in effect reinforce existing preferences and do not expose customers to options that might weaken those preferences.

These platforms not only track the customer’s own preferences in order to guide the customer’s future purchases, but they also aggregate the preferences of other customers. These aggregations may be viewed as the best available data on a matter that is of utility to site users. They are also influential, and can help to perpetuate discrimination. Crowd-sourcing apps, for example, collect information about the safety of neighborhoods based on the reported experiences of users. That information is then consulted by other users when they are determining where to shop, dine, or go to the movies. Sketch Factor is one such app, and it illustrates the problem.59 To the extent that users of Sketch Factor report a neighborhood “sketchy” because of the visibility of a significant number of black people in the neighborhood, the app builds on stereotypes in a way that is likely to be harmful to black-owned businesses in that neighborhood.60 Once a neighborhood is deemed unsafe—however unfairly that conclusion is reached—subsequent activity by consumers is likely to reinforce this conclusion. Fewer whites will visit the neighborhood, triggering further user responses indicating that the neighborhood is unsafe and magnifying the effects of the stereotyping and bias. This phenomenon is called algorithmic discrimination.61 The more ubiquitous and trusted these peer-to-peer apps become, the greater their potential for influencing the reality that customers, through their decisions, reproduce.

This circularity between business practices and customer preferences often make it impossible to know, in any one instance, which drives which. One controlled study reported that fake web users pretending to be male job seekers were more likely to be shown Google ads for high-paying executive jobs than equivalent female job seekers.62 What explains this phenomenon? Do searches by female job seekers create a pattern leading the algorithm to predict lack of interest in certain job categories by women, as compared to their male counterparts? Or do hiring decisions create the pattern that leads the algorithm to predict more frequent matches between high-paying jobs and men, and thus to facilitate these matches? It is hard to know what drives what, when layers of aggregated data launder the reality that the data helps to create and sustain.

Naomi Schoenbaum identifies another concern with these unregulated, online platforms that exemplifies the role that a firm’s business plan can play in reinforcing customer bias. Schoenbaum explains that online sharing platforms are more likely to involve intimate activities (e.g., dating sites) and intimate spaces (e.g., home services like TaskRabbit or Handyman), and to entail the exchange of private information between buyer and seller (e.g., Airbnb). Schoenbaum argues that the greater the intimacy of the transactions conducted in the sharing economy, the more likely and the more explicitly sex bias influences the transaction, and, therefore the more likely, and acceptable over time, sex discrimination becomes.63 The lack of structural features in the sharing economy that have constrained discrimination in the traditional economy, like Equal Employment Opportunity procedures and face-to-face contact between employees and managers, makes the situation worse.64

In sum, it is wrong to assume, in either the traditional economy or the sharing economy, that discrimination by customers and firm practices are distinct from one another. The efficacy argument for regulating firms assumes both that firm practices operate largely independently of customer biases, and that firms have the capacity to stop discrimination by simply not discriminating themselves. This is not so. Firm and consumer preferences are interactive, cumulative, and often mutually self-justifying. The question, examined later in Part IV, is what the law should do about it.

B.     Privacy and Individual Autonomy

1.     The Privacy and Individual Autonomy Rationale

Another potentially powerful explanation for why the law focuses on firm behavior and not customers is the societal commitment to privacy and individual autonomy. This explanation is founded on a commitment to the view that, however undesirable an individual customer’s prejudices may be, they are prejudices that an individual in a free society has a right to have. Although the aim of anti-discrimination law is to change how individuals act, not what they believe, the privacy rationale views decisions about what a customer buys, with whom she dines, and what entertainment she pursues as so basic to the customer’s sense of private identity—including the highly protected domains of her religion and her politics—that they should be made without the interference of the state.65

Public accommodation anti-discrimination statutes already recognize the privacy rationale. As a matter of constitutional law, the duty not to discriminate set forth in these statutes does not apply if a business, club, or other operation is sufficiently “private” or intimate to implicate members’ constitutionally-protected rights of freedom of expression and association.66 These rights are especially strong when the organization has a political or religious interest, and are even more strenuously protected when the expressive interest of the organization is exclusionary and unpopular. A white male supremacist group like the American Front may exclude women and minorities; the Rotary Club may not.

Furthermore, associational rights include the freedom to engage in discriminatory behavior in private spaces—freedom that is not allowed in more public settings.67 Accordingly, although under federal law it is illegal to discriminate in the sale and rental of housing, there is an exemption from anti-discrimination rules for units in small, owner-occupied buildings.68 Consistent with the perceived difference between the public and the private, the exemption is lost if the landlord publicly advertises the availability of the unit on a discriminatory basis.69 These details illustrate how, despite the constitutional importance of privacy and autonomy, they are neither absolute nor self-executing, as the next section further explores.

2.     Limitations on the Privacy Rationale for Regulating Only Discrimination by Firms

The broadly shared expectation of a privacy and autonomy right of individuals to make their own market decisions is deep and strong, but it bears a closer look in the broader historical context of anti-discrimination laws. Race and gender theorists have long observed that society deems an action private or public not because of its inherent boundaries, but on the basis of society’s evolving values.70 At any particular regulatory moment, the zone of privacy or associational interests that society decides to protect reflects a comparative weighing of competing interests. In the 1950s, whether racial integration was important enough to justify intruding into the individual’s private sphere of association was an open question. At the time, as respected a scholar as Herbert Wechsler could write that “integration forces an association upon those for whom it is unpleasant or repugnant,” and that “[g]iven a situation where the state must practically choose between denying the association to those individuals who wish it or imposing it on those who would avoid it,” allowing people to avoid repugnant associations could well be the value best supported by “neutral principles” in the Constitution.71  Robert Bork was less guarded when he identified as a principle of “unsurpassed ugliness” the view that “if I find your behavior ugly by my standards, moral or aesthetic, and if you prove stubborn about adopting my view of the situation, I am justified in having the state coerce you into more righteous paths.”72

Judge Bork’s views had considerable support in 1963, but these opinions would garner little agreement today if raised as an argument against public accommodations law. Likewise, domestic abuse—once viewed as a private matter into which the state should not intrude73—today is considered an appropriate sphere of state interference.74 Conversely, courts recently have expanded privacy protections in some matters once thought suitable for public regulation, including campaign finance and gun control.75 For better or worse, the law about what decisions belong to individuals and not the state has changed, and continues to evolve.

Looking back over the zones of privacy and autonomy that society used to take for granted, it should be clear that, in the context of customer discrimination, it is not enough to assert that individual privacy and autonomy are important interests. It must also be shown that they are more important than the interests with which they compete, including non-discrimination goals. When society prohibits markets in heroin and body parts, it infringes on people’s right to choose to participate in these markets in order to advance a more important interest in protecting its citizens. The question, therefore, is what value society now places on its commitment to anti-discrimination in relation to the competing values relating to privacy and autonomy.

In legal debates about discrimination, there has been no serious attempt to weigh the strength of society’s present interest in preventing customers from discriminating, in its various forms, against society’s present commitment to personal privacy and autonomy for its citizens. Without this inquiry, the privacy and autonomy rationales are arguments, not justifications. Such inquiry requires attention to context. Some types of customer discriminations are not as harmful as others, and some limits on customer discrimination are more intrusive than the alternatives. The relevant context includes the fact that today’s online platform economy already represents a massive voluntary and unregulated “sharing” of personal information, including physical locations, pictures, tastes, travel patterns, work hours, passport and drivers’ license data, and financial details. Given the many uses to which this information is put, it cannot be assumed, categorically, that this information should not be used, even a little bit, to curtail race and gender discrimination.

How to prioritize the competing interests is considered in more detail in the next sections. The point, here, is that any line drawing in this area reflects some set of priorities. While the privacy rationale offers some explanatory power for why the law does not regulate the exercise of customer bias, it begs the question it purports to answer: is the reduction of race and gender discrimination an important enough goal to consider some limits on customer choice to help achieve it?

III.     Some Distinctions

Not all forms of customer discrimination are equally harmful and unacceptable. There are differences between biased choices, both in kind and in degree, that may determine the optimal legal approach for addressing discrimination by customers. Some discrimination by customers is clearly harmful and not justified by competing values; this discrimination warrants the greatest attention. Other kinds of discrimination might appropriately be viewed as acceptable and even desirable—ones with which the law should not attempt to interfere. Still other forms of customer discrimination are harmful but nonetheless should be left alone because of the costs of trying to prevent them.

Sorting customer discrimination into these categories is a fraught task and necessarily tentative. Our suggestions reflect our own values, without the benefit of the debate we hope this Essay will stimulate. In this section, we sort some examples into these categories, at least tentatively, in an effort to provide a fuller context for analyzing appropriate responses to discrimination by customers.

A.     Acceptable (Even Desirable) Customer Discrimination

Some customer discrimination might be considered tolerable, even desirable, because it helps to reverse historical patterns of discrimination that subordinate people based on a protected characteristic, like race or sex. Customers sometimes discriminate by race or gender, in concert, to promote non-discrimination goals. The bus boycott by blacks in Montgomery, Alabama in the mid-1950s was instrumental in creating awareness of racial segregation in the South and, ultimately, in stimulating the civil rights movement. A “girlcott” started by teenagers in 2005 brought attention to Abercrombie & Fitch’s sales and marketing strategies that contributed to unhealthy body images for girls.76 Customers have also boycotted Russian vodka to try to promote the change of Russia’s anti-gay laws.77 Today, numerous consumer campaigns are directed at helping black-owned businesses and to provide opportunities that have long been disproportionately denied to them. BLACKOUT is an international movement designed to encourage customer to redirect their spending and investments to black-owned companies.78 Some brands, like FUBU (“For Us By Us”) are designed to draw customers seeking to keep their business within the black community.79 Even the federal government, acting as customer, engages in discriminatory purchasing decisions when it imposes minority set-aside requirements on federal contractors.80 A truly “color-” or “gender-blind” approach to combatting customer discrimination would not distinguish these instances from the rest. We would allow these practices, however, because they diminish gender and racial inferiority and subordination rather than contribute to it.81

 B.     Regretfully Allowed Customer Discrimination

Other customer discrimination may be undesirable but outweighed by the customer’s interest in privacy and personal autonomy. It is not an easy matter to decide what belongs in this category, but online dating sites might be the best example. Who someone dates is a very personal matter, integral to an individual’s social and sexual identity.82 Who to date is not simply a private choice, but it concerns activities that often take place in private spaces to which courts have afforded special protection.83 Whether or not sexual desire is as inherent to one’s identity as is commonly assumed, the fact remains that the freedom to make one’s own choices with respect to one’s sexuality, family, and procreation is one of this nation’s most protected constitutional rights.84 It seems appropriate to grant this same freedom in market transactions. If individuals can apply their own selective criteria in choosing with whom to dance in a bar—hardly something one can imagine being the subject of state regulation—it seems reasonable that they also be able to do so when they access a commercial dating site.85

This concession to personal privacy and choice in the online dating world is not without cost, from an anti-discrimination standpoint. Race selection in dating helps to maintain the distinctions upon which racial hierarchy is based and perpetuates current biases in dating markets, especially the bias against black women.86 The exercise of these biases, though, seems best understood as a regretful consequence of living in a free society. Few people, even if they applaud the goal of race-neutral dating and follow that goal in their own personal dating decisions, would be willing to sacrifice their own ability to make unfettered decisions involving their sexuality in order to advance society’s broader goal of anti-discrimination. For this reason, it seems doubtful that any amount of debate would likely lead to the regulation of online dating and sex sites like Match.com, Peppr and Tinder in order to prevent customers from making choices based on race or gender preferences.

A similar example of customer discrimination that we would deem regretfully allowable is the choice of a roommate. Again, this kind of choice affects both intimate activities and intimate spaces that we ordinarily protect from public intrusion. In its 2012 holding that it is not a violation of the California Fair Employment and Housing Act for Roommate.com to allow its users to select roommates based on race, the Ninth Circuit Court of Appeals reasoned:

The home is the center of our private lives. Roommates note our comings and goings, observe whom we bring back at night, hear what songs we sing in the shower, see us in various stages of undress and learn intimate details most of us prefer to keep private. Roommates also have access to our physical belongings and to our person.87

As with dating sites, a hands-off approach to roommate discrimination is likely to disproportionately affect members of racial minority groups, given the disproportionate bias they face in other settings. At the same time, the privacy considerations attending one’s private living space are compelling enough to make forced anti-discrimination standards impractical. The same reasoning would not apply to colleges and universities when they deliberately design roommate matches for the educational purpose of enlarging students’ horizons.88

Likewise, it is hard to imagine a direct ban on race matching in the sperm donor market.89 This choice also carries the potential of harmful discrimination, yet it goes to deeply intimate decisions about who might procreate, and how, along with deeply personal interests in controlling one’s own self-definition.

There may also be personal services that are so intimate in nature and space that they, too, should be largely protected on privacy and autonomy grounds. Massage therapists and psychiatrists may fit this category. A massage therapist’s services, arguably, cannot be effectively performed where the client does not feel comfortable, relaxed, and safe from sexual threat. The same reasoning does not apply to preferences that result from stereotypes linking sex to competence, such as hospital patients’ preferences for female nurses or male doctors.90 Others might classify these cases differently91 but, as a general matter, the goal should be to respect customer choices in especially personal settings while putting continued pressure on the stereotypes that influence those choices in a biased way.

C.     Disapproved Customer Discriminations

The third category are those customer choices that cause harm to others and do not as seriously implicate the kinds of privacy interests discussed above. In this category we would include race and gender bias by customers when they avoid store clerks on account of their race and gender or evaluate them in a discriminatory way. We would also include race and gender bias in tipping, buying an iPod, or renting weekend accommodations. Limitations in these instances infringe on customer choice to some degree, but do not intrude on particularly intimate spaces and activities. In this regard, these actions are more like political campaign contributions, bribery, or buying drugs—activities which the state regulates from the customer side despite the intrusion of that regulation on individual choice—than like dating sites and massage therapists.

At present, as regards to sex discrimination in the employment sphere, the line between permitted customer discrimination and disapproved discrimination is policed through Title VII’s BFOQ exception, discussed above.92 It is a line that has evolved over time, as attitudes about discrimination have evolved. At one time, case law held that sex was a BFOQ in a wide variety of prison and medical facility cases.93 Courts have pulled back on the exception,94 and today these cases are comparatively rare. So, too, as customer stereotypes diminish over time, the kinds of discriminatory preferences that are deemed necessary to protect customer privacy interests should also shrink.

Some cases are truly borderline and will be especially controversial. Take choices about one’s doctor, which many patients consider to be highly personal. This choice can be influenced by race and gender stereotypes—that white male professionals are more competent than their female or minority counterparts—and allowing sex or race selection reinforces these stereotypes and deprives women and minorities of professional opportunities. To be sure, patients may have a higher comfort level with doctors who reflect their stereotypes about competence, which may produce better outcomes.95 Yet the same thing could be said about other professional relationships in fields such as the lawyer–client relationship, in which we do not allow sex and race discrimination.

At the same time, some reasons for wanting to choose one’s own doctor are in line, at least somewhat, with anti-discrimination goals. After a history of male control over women’s reproductive lives, for example, many women prefer female gynecologists, and believe that they receive better care from them.96 Similarly, after a history of subpar medical care, many black patients prefer black doctors, whom they believe are less likely to discriminate against them.97 Research shows, in fact, that because of discrimination in the provision of medical care, sex-matching or race-matching may produce higher objective levels of medical care.98

These are tough cases requiring especially creative strategies—strategies that might both take account of the “ordinary” case in which customer discrimination should be prevented, if possible, and cases in which anti-discrimination goals might be best met by allowing sex- or race-based decisions by customers. Society has largely avoided the questions that might give rise to the development of those strategies. This Essay asks these questions and, in the next Part, attempts to provide a possible framework for resolving them.

IV.     An Approach to Customer Discrimination

It bears repeating that existing law already restricts customer preference to a certain extent. By making it illegal for employers, landlords and other providers of public accommodations to discriminate based on race and gender, the law indirectly limits customer choices about who they work with and buy from.99 But discrimination persists that the law does not even try to stop. In this Part, we turn to the regulatory analysis. Are there regulatory tools that might curtail this discrimination effectively and without unacceptable infringement on customer privacy and autonomy?

We conclude that, even as to the forms of discrimination society should deem least acceptable, a direct prohibition of discrimination by customers is not the place to start. Instead, we make a modest proposal to enlarge slightly the obligations that firms already have.

A.     The Case Against an Outright Prohibition

The most obvious strategy for reducing customer discrimination would be an outright prohibition, analogous to the prohibition of discrimination by firms in employment and public accommodations. A direct prohibition is attractive because it would acknowledge the harm of customer discrimination, recognize that prohibiting discrimination by firms does not reach all forms of harmful discrimination by customers, and state a public norm against this discrimination. A prohibition against customer discrimination would also take seriously the proposition that customer privacy and autonomy interests are not trump cards that take automatic priority over the competing interests society has in ending discrimination.

Nonetheless, we are reluctant to advocate for a direct constraint against customer discrimination. Our reasons bring us back to the justifications we first examined for refraining from regulation against customers: efficacy and privacy.

In Part II.A, we questioned some assumptions about the efficacy of a firm-only regulatory approach to discrimination, but not the basic proposition that regulating firm discrimination is generally more efficient than regulating discrimination by customers. One additional piece of evidence for this proposition is the disuse of section 1981 of the Civil Rights Act of 1866 to attack discrimination by customers.100 While section 1981 prohibits discrimination on the basis of race in private contracts, there are virtually no reported cases of section 1981 being applied against individual (i.e., non-corporate) customers,101 nor is there any serious discussion in the literature, either academic or practitioner, about the potential uses of section 1981 as a tool against discrimination by customers.102 This suggests, among other things, that most customer interactions are probably too fleeting and lacking in meaningful documentation for any disgruntled seller to chase after a potentially discriminatory customer with a legal action.

While the case for a hands-off approach to customer discrimination is not fully established by the efficacy rationale, a direct prohibition on customer stereotyping would also likely be a regulatory nightmare. The problem is not that we cannot track consumer behavior; most of our consumer searches and purchases are already monitored in the private sector, and the collected information, such as individual “likes” and “dislikes” is traded in the market.103 The problem is, rather, determining what constitutes discrimination—would it be one act, or many? How many? Would discriminatory intent need to be shown (if so, how?), or would a pattern of discriminatory decisions be sufficient? How convincing would that pattern have to be? Would a pattern of discrimination be shown by successive acts against different sellers, or would there need to be a sufficient number of discriminatory decisions harming the same seller? What kind of harm, if any, would have to be shown?

There is the further matter of customer privacy, examined in Part II.B. Even if a generalized interest in privacy does not justify the absence of all legal strategies to reduce discrimination by customers, it suggests the desirability of a protective zone within which customers can make buying decisions, at least certain kinds, without the threat of legal action. This is a strategic concern, as well as a legal one. Behavioral research has shown that when legal coercion deeply transgresses people’s sense of themselves, they respond defensively with resentment and defiance, rather than acceptance.104 Since the most meaningful progress toward non-discrimination entails changing the underlying attitudes on which discriminatory actions—intended or not—are based, this kind of response is especially counter-productive. Having to pay bridge tolls or meet carbon emission standards is one thing; these are annoyances (even sometimes expensive annoyances) that generally do not challenge a person’s sense of identity, self-definition, or feeling of personal freedom. Constraints on to whom to rent an extra bedroom, how much to tip a restaurant server, or what parameters one can set on an online dating service, are more personal. While strong, normative messages disapproving of discrimination are essential to reshaping norms,105 indirect strategies may be more effective than coercive rules in transmitting those norms.

B.     A Modest Proposal

Consideration of regulatory alternatives brings us back to the firm, the party upon whom society, for good reasons, already relies to reduce discrimination. As noted above, firms, particularly large and well-resourced ones, collect enormous amounts of information about their customers.106 They know their customers’ preferences and buying habits. They have the ability to structure the choices they give customers, influence their preferences and habits, and distribute the costs of discrimination that they are unable, or not required, to eliminate.107

With these factors in mind, we propose that entities that already have a legal obligation not to discriminate—i.e., employers of a certain size, educational institutions, and entities subject to public accommodations laws—also should have an explicit obligation to curtail and not to facilitate discrimination by their customers, and to refrain from discrimination when they, themselves, act in the role of a customer.

The proposal has features that address each of the concerns spelled out in the prior sections of this Essay. First, it imposes duties only upon those who already have an anti-discrimination duty—that is, firms of a certain size. Our analysis of the reasons society does not directly regulate customer choice concluded that, while these reasons do not seem strong enough to abandon an appropriate societal concern with customer discrimination, they are practical realities: it is generally easier and cheaper for the state to regulate firms than customers, and direct regulation against customers will both interfere with the important value of customer choice and be perceived as so offensive to privacy and autonomy values as to be counterproductive to the kind of fundamental changes in people’s preferences necessary to bring about a more open and inclusive society.

By avoiding a direct obligation on the customer, our suggestion blunts the force of these concerns and realities. The proposal requires only those already having an anti-discrimination obligation to avoid facilitating the kind of behavior by others that they are not allowed to engage in themselves. It recognizes that it is easier to regulate firms than individuals, without giving up on the goal of inhibiting customer discrimination. In concession to privacy concerns, the proposed rule does not impose any direct obligations against customers, nor does it interfere directly with those choices. The rule accommodates an exception along the lines of the BFOQ exception to Title VII in situations in which the most compelling instances implicating customer privacy are at risk, although its presumed effect would be to gradually reduce customer bias, and thus to decrease the need for the exception. By not involving the state directly in monitoring customer behavior, the indirect approach avoids the practical enforcement difficulties associated with that monitoring. It also minimizes concerns about backlash arising from state-initiated regulation, relying instead on firms, who make all other business decisions, to find the means to inhibit discrimination by their customers.

With these general advantages, questions remain as to the proposal’s scope and detailed application. The next section expands these details.

C.     Implications of the Proposal

As noted above, anti-discrimination law already, to a certain extent, requires businesses to act in ways that prevent customers from exercising their discriminatory preferences.108 Despite some favorable authority along these lines, however, current law generally does not hold firms accountable for customer discrimination that their own practices allow and that they could prevent. Laws prohibiting disparate impact discrimination in the employment arena,109 which are directed against rules and practices that have a disproportionate impact against a protected class,110 might have been useful to this end. Unfortunately, courts have thrown up roadblocks to disparate impact claims. For example, courts have held that plaintiffs must identify a particular practice that causes the disparate impact,111 which can be next to impossible to do in the case of customer discrimination, given the often interactive nature of decisions that no one party controls.112 Courts also have been, by and large, unwilling to consider non-statistical showings of impact or to credit broader societal data about the effects of a business practice.113 The proposed rule would override these limitations when customer bias causes harm and there are reasonable ways businesses could change their practices to prevent that harm.

An advantage of the proposal is that it reaches firms like Amazon, Uber, and Airbnb that are primarily intermediaries of the transactions of others. Because of their enormous data-gathering capacities, these firms are even better equipped to affect their customers’ behavior than traditional firms.114 Today, they have no such responsibility.115 Under the proposal, these firms would now be responsible for not only the treatment of their own employees and the public they serve, but also for the ways in which their platforms facilitate discrimination by their customers.

There are a number of ways firms can respond to the obligation imposed by the proposed rule. One easy, albeit minimal, step would be to proactively declare that they will not accede to the discriminatory preferences of their customers, and to demand a promise from their users not to discriminate.116 This kind of policy may be difficult to enforce, but nonetheless would help to strengthen a norm against discrimination by customers. Corporate self-identification with anti-discrimination goals could become as standard as the current practice of announcing oneself to be an equal opportunity employer.

So could policies intended to put teeth in these announcements. Some online companies like Google already have policies limiting the kinds of information their users can demand from each other.117 Airbnb has recently announced a host of reforms, including promoting the use of Instant Book, which allows qualified guests to book without prior host approval, making guest photos less prominent in the booking process (on an experimental basis), blocking a booking after a host has previously denied the booking to a qualified guest, and instituting a program called Open Doors, to help guests find books when they have faced discrimination in the booking process.118 Such policies both reduce opportunities for discrimination, and the costs of discrimination to its victims. Businesses can use these kinds of policies to brand themselves in a positive way. Early movers, even if they find themselves at a competitive disadvantage in the near term, might be expected to gain reputationally in the long run.119 Indeed, businesses are likely to conclude that enacting such policies voluntarily benefit them more than waiting for litigation,120 or public outcry.121

Responding to consumer bias, in some contexts, gives businesses a chance to iterate and reinforce their corporate values. Fox News arguably did this when it stood up to Donald Trump’s demands that Megyn Kelly be dropped as a moderator for one of the Republican presidential candidate debates. Trump’s demands were in retaliation for Kelly challenging Trump on his sexist views at a prior debate122 and, without Fox’s tough stand, might have damaged Kelly’s future career opportunities as well as chilled the behavior of other (female) journalists. The incident was unusual for its degree of public salience, but illustrates the kind of measures a firm can take to prevent its customers from undermining the opportunities of its employees, as well as to contribute to positive anti-discrimination norms.123

The proposed rule would also pressure businesses to rethink branding and advertising decisions that reinforce discriminatory customer preferences. This includes advertisements that encourage customers to think of women as sex objects, or as less competent than men. Under the proposal, a sports franchise should rethink a name, or mascot, that disrespects women or minorities; it should refuse, for example, to allow its fans to attend games in “redface”124 or other garb that, while appealing to customer biases, shames or humiliates members of a historically disadvantaged minority.

Not all available measures would necessarily be required by every firm, even in the case of discrimination by customers that falls into the least acceptable category. The proposed rule would only require firms to take reasonable steps calculated to end the harmful effects of discrimination by its customers. Yet, while firms have some discretion, there are some business practices that will be plainly prohibited by the proposed rule. For example, the proposed rule would make plain that firms cannot have hiring policies that assume that members of the public prefer to deal with people of their own race. Because of the direct impact of these policies on employees, some courts have interpreted existing employment discrimination law to outlaw them. The Eleventh Circuit Court of Appeals held, for example, that a black woman who was let go by a telephone marketing company stated a prima facie case of race discrimination under section 1981 based on the company’s practice of “race-matching” black callers and black scripts with phone numbers in black calling areas and white callers and scripts with white calling areas.125 In another case, the Seventh Circuit Court of Appeals reversed summary judgment against a black employee suing for race discrimination based, among other things, on the fact that the employer assigned black salespersons to deal primarily with black accounts and white salespersons to serve white accounts.126 The proposed rule makes these practices more clearly illegal.

Another practice that presumptively violates the proposed rule is reliance on evaluation tools through which biased customers are able to influence the employment prospects of women and minorities. Customer evaluations are notoriously biased along race and gender lines.127 Teacher evaluations by students, likewise, have been shown to reflect significant biases against women and minorities.128 Under this proposal, firms and educational institutions should not rely on these evaluations unless they find a way to negate that bias.129

As for discrimination by customers that should be discouraged but not necessarily banned, some of the best strategies will allow the firm to impose the cost of that discrimination on the customers themselves. The model would be: (1) alcohol and tobacco taxes, by which society discourages socially undesirable behaviors by shifting the negative externalities of those behaviors to those engaged in them; and (2) managed care health plans, in which participants have to pay more to see, say, an out-of-network doctor than one covered by the plan. Again, the proposed rule would not necessarily require all firms to institute the same policies, but where customer bias affects the opportunities of its employees, it would obligate companies to find ways to reduce or eliminate that harm. Uber, for example, might assess a fee on customers who systematically reject dark-skinned drivers with foreign-sounding names. If it did this, the proceeds could be redirected toward the population of drivers negatively affected by such systematic rejections. Other approaches might be appropriate where customer preferences would otherwise impose costs on other customers. Airlines asked to accommodate ultra-Orthodox Jewish men whose religious beliefs forbid them from sitting next to a woman on an airplane,130 for example, could place the costs of those accommodations on the men themselves rather than on female passengers, by requiring the male passenger to make prior arrangements or even to buy an extra (empty) seat.131

Another structural approach is for firms to bundle certain goods and services to reduce the harmful effect of discriminatory customer preferences. One example is tip-pooling in the restaurant business. Tip-pooling levels out the discriminatory effects of race and gender bias in tipping by spreading those effects across all servers. As a result, no one set of servers, defined by race or gender, is disproportionately harmed by customer bias. Under the proposed rule, a restaurant would have an obligation to review a tipping policy that easily lent itself to race or gender discrimination and, unless it discovered that no such discrimination existed in its own establishment, move to a system like tip-pooling that mitigated the discriminatory consequences.

Customers tend to become accustomed, and adapt, to the structure of a market. To take a simple example, law students accept the fact that the professors of the classes to which they are assigned in their required 1L classes represent, usually, a demographic cross-section of the available professors. Few students object to what is, in effect, the bundling that occurs, even though it reduces their choices. Similarly, clients typically accept the diverse work teams that firms establish to provide professional services; in addition to its other advantages,132 this bundling practice both makes it more difficult for clients to discriminate, and creates the working relationships in which prejudice may be reduced.133

Pushing this strategy into less familiar territory, colleges and universities might do well to bundle tickets for men’s and women’s sports teams. Many colleges and universities already prop up the market for their less popular sporting events (say, non-conference games) by bundling tickets to those events with tickets that are more highly sought after (conference games). In the professional world of sports, such as the U.S. Open and other professional tennis events, tickets to men’s and women’s events have been bundled for decades.134 Bundling tickets for men’s and women’s college team competitions would not require fans to attend games they do not wish to attend. It would, however, even out opportunities for women and men in sports, allocate the costs of equalizing men’s and women’s sports to those with discriminatory preferences, and, as would appear to have been the case in professional tennis, induce some fans to develop a taste for women’s sports.

There are undoubtedly other strategies that firms and establishments of public accommodation could use to discourage discrimination by their customers, situate the cost of that discrimination on those engaged in it, or spread that cost more evenly across society. As companies with customers who make choices reflecting race or gender bias are required to answer for their role in this bias, the proposed rule will encourage the development of such strategies.

V.     Conclusion

The direct, measurable effects of the proposed rule, admittedly, are not likely to be large. While some changes in a firm’s practices will be compelled by the proposed rule, the rule allows firms to choose among reasonable measures to reduce discrimination by customers. Perhaps the biggest gap in the proposed rule is that it will have no effect on businesses that customers avoid because of race or gender bias against the business. To address this problem would require a direct ban on customer behavior that, for all the reasons given in this Essay,135 would likely be impractical and ill-advised.

Despite these limitations, the proposed rule requires firms to do something meaningful to ensure that the bias of their customers does not disadvantage some employees. The measures chosen should make discrimination more difficult or allocate its costs more fairly. Just as importantly, these measures can have a positive influence on societal norms. Research suggests that the normative information conveyed by the law and responses to it can be important in enabling individuals in society to coordinate their actions around socially beneficial actions.136 The fact that a robust norm against discrimination has been built up over a half century in the contexts of employment and public accommodations may make the task of extending that norm to individual customers easier. There can be little doubt that anti-discrimination laws have had a significant impact on workplace fairness and in access to public accommodations, despite the fact that plaintiffs in discrimination cases are more likely to lose their lawsuits than plaintiffs in any other category of civil litigation.137 The law prohibiting race and sex discrimination in choosing a jury138 is almost impossible to enforce,139 and yet the law marks the underlying discriminatory behavior as wrong, in part to signal society’s disapproval of discrimination and to build social norms accordingly.140 The proposed rule has the potential to make this same kind of contribution to anti-discrimination norms. The state can help reinforce these same norms through public service campaigns attacking discrimination with the same vigor it has used to attack smoking, littering, and drugs.

Once discrimination by customers is fully understood in relation to the firm behavior with which it is entwined, it no longer seems possible to view it as entirely outside the law’s concern. Customer discrimination harms some people in the same way as discrimination by firms, if fewer at a time. There is at present no law that tells customers, or the businesses they patronize, that customer discrimination is wrong. The proposed rule would change this. It would recognize the harm of discrimination by customers and adopt a norm against it. It would impose an obligation on firms that takes account of the concerns that make regulation in this area such a sensitive matter, and of the different contexts that makes some forms of discrimination by customers more important to stop than others.

The proposed rule may seem, at once, both too incremental and too extreme. This is to be expected. Every extension of anti-discrimination law takes some getting used to, but we have tried to be realistic in terms of what might be politically feasible and actually work. The proposed rule is designed to stimulate imaginative thinking about the problem of customer discrimination. In the course of time, the solutions that firms are able to frame will become as second nature as other measures we have come to count on to improve race and gender equity.


  1. [1]. See generally Jennifer L. Doleac & Luke C.D. Stein, The Visible Hand: Race and Online Market Outcomes (Jan. 25, 2015) (unpublished manuscript), http://ssrn.com/abstract=1615149.

  2. [2]. Id.

  3. [3]. See generally Ian Ayres et al., Race Effects on eBay, 4 Rand. J. Econ. 891 (2011); see also Clark Nardinelli & Curtis Simon, Customer Racial Discrimination in the Market for Memorabilia: The Case of Baseball, 105 Q.J. Econ. 575, 578 (1990) (examining why collectible cards featuring prominent African-American baseball players were valued far below cards featuring white players).

  4. [4]. See Benjamin Edelman & Michael Luca, Digital Discrimination: The Case of Airbnb.com 9 (Harvard Bus. Sch. Negotiation, Orgs. & Mkts. Unit, Working Paper No. 14-054, 2014), http://
    papers.ssrn.com/sol3/papers.cfm?abstract_id=2377353.

  5. [5]. Id. at 3. Follow-on research showed that Airbnb guests with distinctively black names are roughly 16% less likely to be accepted by hosts than identical guests with white names. See Benjamin Edelman et al., Racial Discrimination in the Sharing Economy: Evidence from a Field Experiment 11–12 (Harvard Bus. Sch. Negotiation, Orgs. & Mkts. Unit, Working Paper No. 16-069, 2016), http://www.benedelman.org/publications/airbnb-guest-discrimination-2015-12-09.pdf; see also Michael Todisco, Share and Share Alike? Considering Racial Discrimination in the Nascent Room-Sharing Economy, 67 Stan. L.R. Online 121, 122–23 (2015) (exploring issue of racial discrimination in room-sharing economy); Audie Cornish, When Personalization Leads to Discrimination On Airbnb, NPR: All Things Considered (Apr. 26, 2016, 4:28 PM), http://www.npr.org/2016/04/26/475773261/when-personalization-leads-to-discrimination-on-airbnb. For the point that hosts and guests who use sharing platforms like Airbnb are all customers, see infra notes 57–62 and accompanying text.

  6. [6]. See Ian Ayres et al., To Insure Prejudice: Racial Disparities in Taxicab Tipping, 114 Yale L.J. 1613, 1616 (2005).

  7. [7]. See, e.g., Matthew Parrett, Customer Discrimination in Restaurants: Dining Frequency Matters, 32, J. Lab. Res. 87, 100–03 (2011) (concluding, based on data collected outside five Virginia restaurants, that customers tip female servers less than men, unless their customers are regular patrons or the service quality is exceptional).

  8. [8]. See Female Client Service Reps Get Lower Scores Despite Better Performance and Experience, ThinkProgress (May 22, 2014), https://thinkprogress.org/female-client-service-reps-get-lower-scores-despite-better-performance-and-experience-dfa457062f5c#.ydju1f46m.

  9. [9]. See generally Sylvia R. Lazos, Are Student Evaluations Holding Back Women and Minorities?: The Perils of “Doing” Gender and Race in the Classroom, in Presumed Incompetent: The Intersections of Race and Class for Women in Academia 164 (Gabriella Gutierrez y Muhs et al. eds., 2012); Lillian MacNell et al., What’s in a Name: Exposing Gender Bias in Student Ratings of Teachers, 40 Innovative Higher Educ. 291 (2015); William Arthur Wines & Terence J. Lau, Observations on the Folly of Using Student Evaluations of College Teaching for Faculty Evaluations, Pay, and Retention Decisions and Its Implications for Academic Freedom, 13 Wm. & Mary J. Women & L. 167 (2006).

  10. [10]. We use the term “firm” to refer loosely to groups of individuals operating as a collective enterprise—usually, but not always, economic—to achieve particular goals. We use the term in this Article interchangeably with company, entity, business and, in some cases, educational institutions.

  11. [11]. Title VII of the Civil Rights Act of 1964 prohibits discrimination by employers with 15 or more employees on the basis of race, color, religion, sex or national origin. 42 U.S.C. § 2000e (2012). Title II of the Civil Rights Act of 1964 prohibits racial discrimination in “any inn, motel, or other establishment which provides lodging to transient guests.” 42 U.S.C. § 2000a(b)(1) (2012). The Fair Housing Act prohibits discrimination based on race, color, religion, sex, familial status, or national origin. See 42 U.S.C. §§ 3601–31 (2012). Examples of state public accommodations laws include Cal. Civ. Code §51(b) (West Supp. 2016); and N.J. Stat. Ann.
    § 10:5-4 (West 2013).

  12. [12]. The exceptions are section 1981 of the Civil Rights Act of 1866, which prohibits race discrimination in private contracts, and section 1982 of the same Act, which prohibits race discrimination in property transactions. 42 U.S.C. §§ 1981–82 (2012). These statutes apply only to race discrimination, however, and have not been applied, in practice, against individual customers. The few cases brought against corporate customers have been easily defended, usually because of an inability to show discriminatory intent. See, e.g., TRI, Inc. v. Boise Cascade Office Prods., 315 F.3d 915, 919 (8th Cir. 2003) (dismissing on the basis of nondiscriminatory reason for defendant refusing to deal with plaintiff); Arduini v. NYNEX, 129 F. Supp. 2d 162, 170 (N.D.N.Y. 2001) (dismissing for lack of allegation of racial motivation); Pippen v. Scales, 822 F. Supp. 305, 311 (M.D.N.C. 1993) (dismissing a claim under section 1981 for failing to allege discriminatory intent). For an exploration of the use of section 1981 by customers when they are discriminated against by firms, see generally Abby Morrow Richardson, Applying 42 U.S.C. 1981 to Claims of Consumer Discrimination (Bepress Legal Series, Working Paper No. 651, 2005), http://law.bepress.com/expresso/eps/651. For some of the difficulties involved in proving a section 1981 claim, see Ian Ayres, Pervasive Prejudice?: Unconventional Evidence of Race and Gender Discrimination 127–36 (2001). For a discussion of the application of section 1982 to home seekers, see generally Lee Anne Fennell, Searching for Fair Housing, 97 B.U. L. Rev. (forthcoming 2017), http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2762026. What the ineffectiveness of these sections might tell us about the efficacy of a prohibition against discrimination directly against customers is explored further infra, Part IV.A.

  13. [13]. Although, theoretically at least, firms may open themselves up to a section 1981 or section 1982 challenge if they refuse to buy from a business on account of race. See supra note 12 and accompanying text.

  14. [14]. Although selling, quite often determined by the amount in possession, is typically considered more serious than buying. For a chart of federal drug possession and trafficking penalties, see Federal Trafficking Penalties, Drug Enforcement Agency, https://www.dea.gov/
    druginfo/ftp3.shtml (last visited Sept. 11, 2016).

  15. [15]. Traditionally, prostitution laws were aimed primarily against the prostitutes. See Sylvia A. Law, Commercial Sex: Beyond Decriminalization, 73 S. Cal. L. Rev. 523, 527 (2000). Led by Sweden, Norway and Canada, there is some movement in the law today toward penalizing only the customer side of the transaction. For a discussion of the Swedish law, see Max Waltman, Prohibiting Sex Purchasing and Ending Trafficking: The Swedish Prostitution Law, 33 Mich. J. Int’l L. 133 (2011).

  16. [16]. See National Organ Transplant Act § 301, Pub. L. No. 98-507, 98 Stat. 2339, 2346 (1984).

  17. [17]. See, e.g., 18 U.S.C. § 201(b)–(c) (2012).

  18. [18]. See infra Part IV.A.

  19. [19]. A new article by Lee Anne Fennell addresses discrimination by home seekers in the fair housing context. See generally Fennell, supra note 12. Another work to note the lack of symmetry in non-discrimination obligations is Tarunabh Khaitan, A Theory of Discrimination Law 199­–200 (2015).

  20. [20]. See, e.g., Ellen Huet, Why Aren’t There More Female Uber and Lyft Drivers?, Forbes (Apr. 9, 2015), http://www.forbes.com/sites/ellenhuet/2015/04/09/female-uber-lyft-drivers (explaining that Uber does not allow drivers to block a certain passenger and does not remove passengers with low ratings, but drivers with low ratings are let go).

  21. [21]. The unique features of the sharing economy raise issues for other areas of local, state, and federal law as well, including health and safety rules, product liability rules, landlord–tenant law, zoning restrictions, tax laws, and labor laws. For an analysis of these and other areas, see generally Vanessa Katz, Regulating the Sharing Economy, 30 Berkeley Tech. L.J. 1067 (2015).

  22. [22]. That the law reflects some assessment of the capacity of businesses to bear the burden of anti-discrimination is apparent in the fact that small businesses do not have the same anti-discrimination obligations as larger ones. Only companies with 15 or more employees are subject to the anti-discrimination prohibitions of Title VII or the Americans with Disabilities Act (“ADA”). See 42 U.S.C. 2000e(b) (2012). The Family and Medical Leave Act (“FMLA”), which requires up to a three-month leave for the case of a new infant or ill family member, only applies to companies who employ more than 50 employees. See 29 U.S.C. § 2611(4)(A)(i) (2012). Likewise, businesses that are small enough sometimes are exempt from public accommodation laws. See infra text accompanying notes 66–69.

  23. [23]. On the ability of “large, sophisticated firms” to “detect and root out internal legal violations—and otherwise alter employees’ and contractors’ behavior—more easily than public authorities or outside private attorneys,” see Brishen Rogers, The Social Costs of Uber, 82 U. Chi. L. Rev. Dialogue 85, 90 (2015).

  24. [24]. For example, after courts disallowed airlines hiring only female flight attendants and imposing unreasonable appearance and weight restrictions on them, the traveling public soon got used to, and came to expect, male as well as female flight attendants, and even female flight attendants who did not conform to traditional criteria for female attractiveness. See Diaz v. Pan Am World Airways, 442 F.2d 385, 389 (5th Cir. 1971); Wilson v. Southwest Airlines, 517 F. Supp. 292, 305 (N.D. Tex. 1981). Cynthia Estlund makes a similar point about the beneficial effects of contact in the workplace. See Cynthia Estlund, Working Together: How Workplace Bonds Strengthen a Diverse Democracy 23 (2003). The literature on the benefits of contact in reducing prejudice is legion. Much of it builds on the “contact hypothesis” developed by Gordon Allport. See generally Gordon Allport, The Nature of Prejudice 281 (1954). For studies linking exposure and contact to the reduction of discriminatory biases, see Katharine T. Bartlett, Making Good on Good Intentions: The Critical Role of Motivation in Reducing Implicit Workplace Discrimination, 95 Va. L. Rev. 1893, 1946–56 (2009). Lee Anne Fennell refers to the possible effect of enlarging one’s horizons and altering one’s biases because of the different choices a person might be influenced to make as the “learning effect.” See Fennell, supra note 12, at 29.

  25. [25]. This was a key insight of Becker’s classic work on the economics of discrimination. See generally Gary S. Becker, The Economics of Discrimination (1957).

  26. [26]. There is an extensive literature on implicit bias. See, e.g., Anthony G. Greenwald et al., A Unified Theory of Implicit Attitudes, Stereotypes, Self-Esteem, and Self-Concept, 109 Psychol. Rev. 3 (2002); Anthony G. Greenwald & Mahzarin R. Banaji, Implicit Social Cognition: Attitudes, Self-Esteem, and Stereotypes, 102 Psychol. Rev. 4 (1995).

  27. [27]. In the Airbnb study, for example, hosts who reject black guests are able to find a replacement guest only 35% of the time. See Edelman et al., supra note 5.

  28. [28]. We have oversimplified this analysis to make the point. It might be rational for a
    profit-driven business to discriminate in order, say, to enhance the camaraderie among its men (by discriminating against women), or save on search costs. On search costs, see generally Shelly J. Lundberg & Richard Startz, Private Discrimination and Social Intervention in Competitive Labor Markets, 73 Am. Econ. Rev. 340 (June 1983). Current law would prohibit such an exclusion. Conversely, customers might have competitive or profit-driven reasons not to discriminate. Our argument here is that the competitive, economic pressures of a profit-driven business provide incentives to refrain from discrimination that are often not present in individual customers.

  29. [29]. Importantly, the sharing economy has made peer-to-peer evaluations ubiquitous.

  30. [30]. See generally Devon W. Carbado & Mitu Gulati, Working Identity, 85 Cornell L. Rev. 1259 (2000) (describing strategies pursued by women and minorities to avoid triggering gender and racial stereotyping by others).

  31. [31]. Coaches of women’s teams are more likely male than vice versa. See Matthew J.X. Malady, Sidelined: Why Are There Still No Women Coaching Men’s Sports, and Why Don’t We Care?, Slate (Sept. 28, 2012, 8:45 AM), http://www.slate.com/articles/double_x/doublex/2012/09/
    female_coaches_why_aren_t_there_more_women_in_charge_of_men_s_teams_.html.

  32. [32]. See 29 U.S.C. § 206(d)(1) (2012).

  33. [33]. See, e.g., Stanley v. Univ. of S. Cal., 13 F.3d 1313 (9th Cir. 1994), aff’d, 178 F.3d 1069 (9th Cir. 1999) (rejecting equal pay claim for an increase of a women’s basketball coach’s salary from $60,000 to $150,000 to match salary of men’s basketball coach, even though she was much more successful in terms of team wins, but whose average attendance at women’s games was 762, as compared to 4,103 for men’s games).

  34. [34]. See generally Equal Emp’t Opportunity Comm’n v. Madison Cmty. Unit Sch. Dist. No. 12, 818 F.2d 577 (7th Cir. 1987) (suggesting in dictum that revenue differentials might justify pay differences).

  35. [35]. The N.C.A.A., for example, justifies paying men’s teams roughly $1.56 million per victory in their national tournament, as compared to nothing for women’s teams because, it claims, men’s games bring in more money. See Andrew Zimbalist, The N.C.A.A.’s Women Problem, N.Y. Times (Mar. 25, 2016), http://www.nytimes.com/2016/03/26/opinion/the-ncaas-women-problem.html. In pro soccer, the top five players on the men’s U.S. national team make an average of $406,000 per year, in comparison to $72,000 per year for members of the women’s team. See Carli Lloyd, Carli Lloyd: Why I’m Fighting for Equal Pay, N.Y. Times (Apr. 10, 2016), http://www.nytimes.com/2016/04/11/sports/soccer/carli-lloyd-why-im-fighting-for-equal-pay.html.

  36. [36]. Deborah Brake, Revisiting Title IX’s Feminist Legacy: Moving Beyond the Three-Part Test, 12 Am. U. J. Gender, Soc. Pol’y & L. 453, 481 (2004); see also Catharine A. MacKinnon, Feminism Unmodified: Discourses on Life and Law 120 (1987) (arguing that “being female and being athletic have been socially contradictory and that being male and being athletic ha[s] been more or less socially synonymous”).

  37. [37]. See James K. Gentry & Raqual Meyer Alexander, Pay for Women’s Basketball Coaches Lags Far Behind That of Men’s Coaches, N.Y. Times (Apr. 2, 2012), http://www.nytimes.com/2012/
    04/03/sports/ncaabasketball/pay-for-womens-basketball-coaches-lags-far-behind-mens-coaches.html (explaining that coaches of men’s teams are paid nearly twice that of coaches of women’s teams and that their pay is increasing at a far higher rate).

  38. [38]. For analysis of the law relating to dress and appearance standards, including standards of beauty, as well as how these standards harm women, see generally, Deborah L. Rhode, The Beauty Bias: The Injustice of Appearance in Life and Law (2010).

  39. [39]. In Jespersen v. Harrah’s Operating Co., Inc., 444 F.3d 1104, 1108–11 (9th Cir. 2006) (en banc), the Ninth Circuit Court of Appeals held that Harrah’s Casino could specify that the women employees in certain job categories should wear face powder, blush, mascara and lip color (“in complimentary colors”) and have their hair “teased, curled, or styled” (and “down at all times”), while men were not allowed to wear make-up or nail polish and their hair could not be below the top of their shirt collars (“[p]onytails are prohibited”).

  40. [40]. The BFOQ exception applies only to sex discrimination, not race discrimination, although there may a kind of de facto BFOQ that operates in some settings, as with police departments who hire minority police personnel specifically to work in minority communities. See generally Kate Manley, The BFOQ Defense: Title VII’s Concession to Gender Discrimination, 16 Duke J. Gender. L. & Pol’y 169 (2009).

  41. [41]. See generally Healey v. Southwood Psych. Hosp., 78 F.3d 128 (3d Cir. 1996) (upholding sex as a BFOQ in hiring counselors at psychiatric hospital); Jennings v. N.Y. State Office of Mental Health, 786 F. Supp. 376 (S.D.N.Y. 1992) (permitting sex as BFOQ in positions at psychiatric hospital involving assistance to naked or partially dressed patients); Jones v. Hinds Gen. Hosp., 666 F. Supp. 933 (S.D. Miss. 1987) (upholding sex-based hiring of male and female nurse assistants and orderlies where job responsibilities included viewing or touching private parts of patients); Brooks v. ACF Indus., Inc., 537 F. Supp. 1122 (S.D. W. Va. 1982) (upholding sex as BFOQ for cleaning men’s bathhouses). But see EEOC v. Hi 40 Corp., 953 F. Supp. 301 (W.D. Mo. 1996) (rejecting sex as a BFOQ in weight-loss centers even though clientele was mostly female and did not want men taking their body measurements); Slivka v. Camden-Clark Mem’l Hosp., 594 S.E.2d 616 (W. Va. 2004) (invalidating hospital policy of hiring only female obstetrics nurses, despite evidence that 80% of patients demanded female nurses and female nurses were needed as chaperones for male physicians).

  42. [42]. Fesel v. Masonic Home of Del., Inc., 447 F. Supp. 1346, 1352 (D. Del. 1978).

  43. [43]. In a different context, the Supreme Court in Dothard v. Rawlinson upheld the exclusion of women from jobs as correctional counselors in maximum-security prisons in Alabama because, as women, they were more likely to stir up prisoners, thus heightening the risk of sexual assaults or other violence. Dothard v. Rawlinson, 433 U.S. 321, 335–37 (1977). As in other BFOQ cases, in Dothard attitudes and prejudices of people outside the challenged employment relationship are treated as givens, and used to justify sex-based practices within that relationship.

  44. [44]. See Naomi Schoenbaum, The Law of Intimate Work, 90 Wash. L. Rev. 1167, 1193 (2015). Vicki Schultz makes the same argument with respect to employees, whose preferences, she argues, are shaped by employer expectations and institutional culture and practices. See Vicki Schultz, Telling Stories About Women and Work: Judicial Interpretations of Sex Segregation in the Workplace in Title VII Cases Raising the Lack of Interest Argument, 103 Harv. L. Rev. 1749, 1815 (1990).

  45. [45]. See Jennifer L. Aaker et al., Nontarget Markets and Viewer Distinctiveness: The Impact of Target Marketing on Advertising Attitudes, 9 J. Consumer Psychol. 127, 137 (2000) (demonstrating that target marketing induces identification by target audiences, especially among distinctive, non-majority groups).

  46. [46]. See The Marlboro Man, AdvertisingAge (March 29, 1999), http://adage.com/article/
    special-report-the-advertising-century/marlboro-man/140170.

  47. [47]. See Lauren Johnson, Abercrombie & Fitch Supports Racism, Sexism and Size Discrimination, Beautiful (May 10, 2013), http://www.beutifulmagazine.com/2013/05/10/abercrombiefitch-supports-racism-sexism-size-discrimination.

  48. [48]. See Barbara Herman, The Unbearable Whiteness of Abercrombie & Fitch: In a Multicultural Era, Can This Branding Still Work?, Int’l Bus. Times (Feb. 28, 2015), http://www.ibtimes.com/
    unbearable-whiteness-abercrombie-fitch-multicultural-era-can-branding-still-work-1830958.

  49. [49]. See, e.g., 10 Recent Racist Ads That Companies Wish You Would Forget, Business Insider, http://www.businessinsider.com/the-10-most-racist-ads-of-the-modern-era-2012-6?op=0#intel-released-an-ad-they-knew-was-racist-7 (last visited Oct. 18, 2016).

  50. [50]. Id.

  51. [51]. See, e.g., text accompanying notes 77–82.

  52. [52]. Wilson v. Southwest Airlines, 517 F. Supp. 292 (N.D. Tex. 1981).

  53. [53]. Id. at 294 (describing ideal representative of the campaign as a “lady” who “is young and vital... she is charming and goes through life with great flair and exuberance... you notice first her exciting smile, friendly air, her wit”).

  54. [54]. Southwest Airlines’ new makeover, recently unveiled, continues to sport a bold, multi-color heart, centered on the belly of its aircraft. See Susan Carey, Southwest Airlines Unveils New Look Echoing Traditional Image, Wall Street J. (Sept. 8, 2014, 1:59 PM), http://online.wsj.com/
    articles/southwest-airlines-unveils-new-look-echoing-traditional-image-1410191427.

  55. [55]. Trade magazines make this clear to their readers. See, e.g., Van Thompson, Racial Discrimination in Advertisements, Ariz. Republic, http://yourbusiness.azcentral.com/racial-discrimination-advertisements-20941.html (last visited Sept. 27, 2016) (“Racism in advertising is not in itself illegal, but that doesn’t mean it’s wise.”).

  56. [56]. See supra Part I.

  57. [57]. Id. There is liability under employment discrimination laws, of course, when the company functions as an employer rather than a transaction facilitator. As others have noted, however, there is a broader trend away from employer-employee relationships and toward market contracting, even without factoring in the phenomenal rise in platforms such as Uber and Airbnb. See generally Matthew T. Bodie, Employees and the Boundaries of the Corporation, in Research Handbook on the Economics of Corporate Law (Claire Hill & Brett H. McDonnell eds., 2012).

  58. [58]. For a discussion of the impact of data collection by intermediaries like Google, see generally Nathan Newman, The Costs of Lost Privacy: Consumer Privacy and the Rising Income Inequality in the Age of Google, 40 Wm. Mitchell L. Rev. 849 (2014).

  59. [59]. See Rich Shapiro, Going to a Sketchy Neighborhood? New App Called SketchFactor Will Tell You, N.Y. Daily News (Aug. 7, 2014, 10:00 AM), http://www.nydailynews.com/news/national/new-sketchfactor-app-shows-streets-safe-article-1.1896091.

  60. [60]. See Emma Cueto, Sketch Factor App Is Racist—Not On Purpose, but That Doesn’t Make It Better, Bustle (Aug. 8, 2014), http://www.bustle.com/articles/35065-sketch-factor-app-is-racist-not-on-purpose-but-that-doesnt-make-it-better. This problem was the premise of a November 2015 episode of The Good Wife, concerning a black-owned restaurant that saw its business decline, presumably because of a mapping app (ChumHum) that rated neighborhood safety by aggregating customer prejudices about unsafe neighborhoods. See Rachel Goodman, The Good Wife Tackles Algorithmic Discrimination. Meanwhile, in Real Life..., ACLU (Dec. 3, 2015), https://www.aclu.org/blog/speak-freely/good-wife-tackles-algorithmic-discrimination-meanwhile-real-life.

  61. [61]. See Claire Cain Miller, When Algorithms Discriminate, N.Y. Times (July 9, 2015), http://www.nytimes.com/2015/07/10/upshot/when-algorithms-discriminate.html; see also Salon Barocas & Andrew D. Selbst, Big Data’s Disparate Impact, 104 Calif. L. Rev. 671, 677–92 (2016) (describing how supposedly neutral data aggregation has the potential to exacerbate discrimination).

  62. [62]. See Amit Datta et al., Automated Experiments on Ad Privacy Settings, 1 Proc. on Privacy Enhancing Techs. 92, 105 (Apr. 2015).

  63. [63]. See generally Naomi Schoenbaum, Gender and the Sharing Economy, 42 Fordham Urb. L.J. (forthcoming 2016).

  64. [64]. Id.

  65. [65]. Adrienne Davis calls this right “decisional privacy.” See Adrienne D. Davis, Regulating Sex Work: Erotic Assimilationism, Erotic Exceptionalism, and the Challenge of Intimate Labor, 103 Calif. L. Rev. 1195, 1237–39 (2015).

  66. [66]. See Bd. of Dirs. of Rotary Int’l v. Rotary Club of Duarte, 481 U.S. 537, 546 (1987) (holding that the Rotary Club is not sufficiently private to be exempt, under the First Amendment, from public accommodation anti-discrimination obligations); NAACP v. Button, 371 U.S. 415, 428 (1963) (holding that a civil rights organization established to protect individual rights from government intrusion has First Amendment rights that protect it from state charges against its attorneys for barratry and champerty).

  67. [67]. Adrienne Davis calls this “spatial privacy.” See Davis, supra note 65, at 1235–37.

  68. [68]. See 42 U.S.C. § 3603(b) (2012).

  69. [69]. See 42 U.S.C. § 3604(c) (2012).

  70. [70]. The literature is massive. See, e.g., Nadine Taub & Elizabeth M. Schneider, Women’s Subordination and the Role of Law, in The Politics of Law: A Progressive Critique 151, 151–57 (David Kairys rev. ed., 1990) (arguing that public and private labels are socially constructed ones); Frances E. Olsen, The Family and the Market: A Study of Ideology and Legal Reform, 96 Harv. L. Rev. 1497, 1563–67 (1983) (criticizing the public/private dichotomy between family and market as based on political and social values, not inherent distinctions).

  71. [71]. See Herbert Wechsler, Toward Neutral Principles of Constitutional Law, 73 Harv. L. Rev. 1, 34 (1959) (suggesting that as between “denying the association to those individuals who wish it” and not “imposing it on those who would avoid it,” the latter would be given higher value). Twenty-five years later, Justice Powell expressed a similar view, when he argued that law firms should be allowed to associate themselves as they wished, even if this meant discrimination based on sex. See Hishon v. King & Spalding, 467 U.S. 69, 79–80 (1984) (Powell, J., concurring).

  72. [72]. Robert Bork, Civil Rights—A Challenge, New Republic, Aug. 31, 1963, at 21, 22; see also Christopher W. Schmidt, Defending the Right to Discriminate: The Libertarian Challenge to the Civil Rights Movement, in Signposts: New Directions in Southern Legal History (Sally C. Hadden & Patricia Minter eds., 2013).

  73. [73]. See, e.g., Frazier v. State, 86 S.W. 754 (Tex. Crim. App. 1905); see also Model Penal Code §213.1 cmt. 8(c) (1980) (Am. Law Inst. 1980) (explaining the need for spousal immunity in rape cases on the grounds that liability would “thrust the prospect of criminal sanctions into the ongoing process of adjustment in the marital relationship”).

  74. [74]. The debate is ongoing about the appropriate level of state interference, given competing goals individual protection and individual autonomy. See generally, e.g., Jeannie Suk, At Home in the Law: How the Domestic Violence Revolution Is Transforming Privacy (2009) (arguing that the breakdown of the boundary between public and private space that the domestic violence revolution helped to accomplish has reduced autonomy and privacy for both women and men); Elizabeth Ben-Ishai, The Autonomy-Fostering State: “Coordinated Fragmentation” and Domestic Violence Services, 17 J. Pol. Phil. 307 (2009) (urging mandatory prosecution in domestic violence cases, but with a loosely coupled set of services to mediate the effects of state intrusion on women’s privacy and autonomy interests); Elizabeth M. Schneider, The Violence of Privacy, 23 Conn. L. Rev. 973 (1991) (arguing that the failure of the state to intervene in domestic violence is itself a form of violence).

  75. [75]. See generally MacDonald v. City of Chicago, 561 U.S. 742 (2010) (holding that gun control regulation violates Second Amendment); Citizens United v. FEC, 558 U.S. 310 (2010) (holding that campaign finance limits violate First Amendment rights of a non-profit organization); District of Columbia v. Heller, 554 U.S. 570 (2008) (striking down D.C. gun control rules). For discussion of the link between gun rights and interests in privacy and autonomy, see generally Darrell A.H. Miller, Guns as Smut: Defending the Home-Bound Second Amendment, 109 Colum. L. Rev. 1278 (2009). See also Robert Weisberg, The Utilitarian and Deontological Entanglement of Debating Guns, Crime, and Punishment in America, 71 U. Chi. L. Rev. 333, 337 (2004) (reviewing Guns, Crime, and Punishment in America (Bernard E. Harcourt ed., 2003)) (identifying the association of the gun with a form of individual autonomy).

  76. [76]. Jay Karlson, 10 Famous Boycotts, Listverse (Sept. 3, 2011), http://listverse.com/2011/
    09/03/10-famous-boycotts.

  77. [77]. See Tansy Hoskins, Can Consumer Boycotts Change the World?, Guardian (Aug. 13, 2013, 7:58 AM), http://www.theguardian.com/sustainable-business/consumer-boycott-change-world.

  78. [78]. See Buy Black Today, Black Bus. Network, http://www.blackbusinessnetwork.com/
    Store (last visited Sept. 12, 2016).

  79. [79]. See FUBU Collection, http://www.fubu.com (last visited Sept.12, 2016).

  80. [80]. See Executive Order 11246, 3 C.F.R. 1964–65 Comp., p. 339.

  81. [81]. See Michael Blake, The Discriminating Shopper, 43 San Diego L. Rev. 1017, 1033 (2006) (proposing that customer discrimination should only be prohibited if it “exacerbate[s] or contribute[s] to a pattern of social marginalization leading to political exclusion and injustice”). This position aligns itself more broadly with a nonsubordination approach to race and sex discrimination. Among classic defenses of the position are Owen M. Fiss, Groups and the Equal Protection Clause, 5 Phil. & Pub. Aff. 107 (1976); Charles R. Lawrence III, The Id, the Ego, and Equal Protection: Reckoning with Unconscious Racism, 39 Stan. L. Rev. 317 (1987); and Reva Siegel, The Supreme Court 2012 Term, Foreword: Equality Divided, 127 Harv. L. Rev. 1 (2013).

  82. [82]. See generally Lawrence v. Texas, 539 U.S. 558 (2003).

  83. [83]. See generally id.

  84. [84]. See generally id.; Griswold v. Connecticut, 381 U.S. 479 (1965); Skinner v. Oklahoma ex rel. Williams, 316 U.S. 535 (1942).

  85. [85]. See generally Elizabeth F. Emens, Intimate Discrimination: The State’s Role in the Accidents of Sex and Love, 133 Harv. L. Rev. 1307 (2009) (arguing that while the state has an important role in facilitating relationships and interactions across lines of race, gender, and disability, individuals should not be directly pressured into crossing these barriers).

  86. [86]. See Ralph Richard Banks, Is Marriage For White People?: How the African American Marriage Decline Affects Everyone 33 (2011) (citing statistics showing that black women have the lowest marriage rates of any group, due in part to a high black incarceration rate and the fact that black men are three times more likely to marry outside of their race than black women).

  87. [87]. Fair Hous. Council of San Fernando Valley v. Roommate.com, 666 F.3d 1216, 1221 (9th Cir. 2012). For an earlier decision to the contrary in the context of two roommates who were looking for a third, see Larrick, FEHC Dec. No. 98-12, 1998 WL 750901 (Cal. Fair Emp. & Housing Commission July 22, 1998).

  88. [88]. Allowing people to discriminate based on race and sex in choosing a roommate also would not prohibit the establishment of a law making it harder to discriminate, such as by prohibiting advertising on a discriminatory basis, as in the fair housing context with respect to owner-occupied small units. See Fennell, supra note 12, at 28–29. Such constraints, though, are more often associated with disapproved customer discrimination, discussed in the next section. An argument can be made, as Fennell does, that roommate discrimination belongs in that category. We do not take that position, but consider it a close case.

  89. [89]. The matter is hardly free from complication. See generally Dov Fox, Racial Classification in Assisted Reproduction, 118 Yale L.J. 1844 (2009); Guido Pennings, The Right to Choose Your Donor: A Step Towards Commercialization or a Step Towards Empowering the Patient?, 15 Hum. Reprod. 508 (2000).

  90. [90]. But see infra text accompanying notes 96–99.

  91. [91]. See, e.g., Schoenbaum, supra note 44, at 1190–96 (resisting any BFOQ exception for personal services).

  92. [92]. See supra text accompanying notes 40–43.

  93. [93]. See supra notes 40–43.

  94. [94]. See generally Breiner v. Nev. Dep’t of Corr., 610 F.3d 1202 (9th Cir. 2010) (holding that sex is not a BFOQ for working in women’s prison); Slivka v. Camden-Clark Mem’l Hosp., 594 S.E.2d 616 (W. Va. 2004) (holding that sex is not a BFOQ for hiring only female obstetric nurses, despite evidence that 80% of patients demanded female nurses).

  95. [95]. See generally Kimani Paul-Emile, Patients’ Racial Preferences and the Medical Culture of Accommodation, 60 UCLA L. Rev. 462 (2012) (arguing that doctor–patient race- and gender-preference accommodations serve therapeutic goals, and also lessen the problem of discrimination against minorities in the provision of health care).

  96. [96]. See Elizabeth A. Howell et al., Do Women Prefer Female Obstetricians?, 99 Obstetrics & Gynecology 1031, 1031 (2002) (showing that based on convenience sample, 34% of women preferred a female gynecologist); Julie Schmittdiel et al., Women’s Provider Preferences for Basic Gynecology Care in a Large Health Maintenance Organization, 8 J. Women’s Health & Gender-Based Med. 825, 825 (1999) (showing that 52.2% of women in a large health maintenance organization preferred female gynecologists).

  97. [97]. See Jennifer Malat & Mary Ann Hamilton, Preference for Same-Race Health Care Providers and Perceptions of Interpersonal Discrimination in Health Care, 47 J. Health & Soc. Behav. 173,
    183–84 (2006) (showing that racial minorities are generally more likely to seek a doctor of the same race, possibly because of general perceptions of discrimination in the provision of health care and the belief that physicians are less likely to discriminate against patients of their own race); Aasim I. Padela et al., Patient Choice of Provider Type in the Emergency Department: Perceptions and Factors Relating to Accommodation of Requests for Care Providers, 27 Emergency Med. J. 465, 466 (2010) (citing a convenience study of physicians concluding that 24% of patients believe that they get better care from a physician of the same gender, and 32% of patients believe they get better care from a physician of the same race).

  98. [98]. See, e.g., Marianne Schmid et al., Racial Differences in the Surgical Care of Medicare Beneficiaries with Localized Prostate Cancer, 2 JAMA Oncology 85, 85 (2016) (showing that in the treatment of localized prostate cancer, black patients had a seven-day treatment delay and were less likely to undergo lymph-node dissection).

  99. [99]. See supra Part II.A.1.

  100. [100]. Civil Rights Act of 1866, 42 U.S.C. § 1981 (2012).

  101. [101]. See supra note 12.

  102. [102]. For consideration of the application of section 1982 to home seekers, see Fennell, supra note 12.

  103. [103]. Companies like Alibaba have so much information about their customers that they appear to be using that information to evaluate their customers (good, bad, etc.) so as to be able to determine how much credit to give them. See Neil Gough, Alibaba Data to Help Provide Credit Ratings, N.Y. Times: DealBook (Jan 28, 2015, 3:33 AM), http://dealbook.nytimes.com/
    2015/01/28/alibaba-creates-a-consumer-credit-rating-service.

  104. [104]. See generally Michelle M. Duguid & Melissa C. Thomas-Hunt, Condoning Stereotyping?: How Awareness of Stereotyping Prevalence Impacts Expression of Stereotypes, 100 J. Applied Psychol. 343 (2014) (describing experiments suggesting that knowledge about stereotypes actually increased stereotype-consistent behaviors); see also Bartlett, supra note 24, at 1936–41 (summarizing the research); Adam Grant & Sheryl Sandberg, Opinion, When Talking About Bias Backfires, N.Y. Times (Dec. 6, 2014), http://www.nytimes.com/2014/12/07/opinion/sunday/adam-grant-and-sheryl-sandberg-on-discrimination-at-work.html.

  105. [105]. See Bartlett, supra note 24, at 1934–35.

  106. [106]. See supra notes 23, 58–61 and accompanying text.

  107. [107]. See Rogers, supra note 23, at 90.

  108. [108]. See supra Part II.A.1.

  109. [109]. Title VII and the Fair Housing Act impose such a prohibition. See Title VII, 42 U.S.C.
    § 2000e–2(k)(1)(A)(i) (2012) (defining unlawful employment practices based on disparate impact); Fair Housing Act, 42 U.S.C. § 3601 (2012).

  110. [110]. For scholarly works discussing disparate impact discrimination, its relation to disparate treatment discrimination, and the barriers to its success as a robust tool to reduce discrimination, see generally George Rutherglen, Disparate Impact, Discrimination, and the Essentially Contested Concept of Equality, 74 Fordham L. Rev. 2313 (2006); Charles A. Sullivan, Disparate Impact: Looking Past the Desert Palace Mirage, 47 Wm. & Mary L. Rev. 911 (2005); see also generally Tristin K. Green, The Future of Systemic Disparate Treatment Law, 32 Berkeley J. Emp. & Lab. L. 395 (2011) (articulating a theory of systemic discrimination that includes disparate impact cases within disparate treatment law).

  111. [111]. See, e.g., Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 658 (1989) (“As a general matter, a plaintiff must demonstrate that it is the application of a specific or particular employment practice that has created the disparate impact under attack.”); Ilhardt v. Sara Lee Corp., 118 F.3d 1151, 1156 (7th Cir. 1997) (“To prove a disparate impact claim, Ilhardt must prove by a preponderance of the evidence that a particular employment practice had an adverse disparate impact on women on the basis of their sex.”).

  112. [112]. See supra Part II.A.2.

  113. [113]. See Ilhardt, 118 F.3d at 1156–57. More generally, and for other cites, see Joanna L. Grossman, Pregnancy, Work, and the Promise of Equal Citizenship, 98 Geo. L.J. 567, 615–18 (2010).

  114. [114]. See Rogers, supra note 23, at 90.

  115. [115]. See, e.g., Chi. Lawyers’ Comm’n v. Craigslist, Inc., 519 F.3d 666, 671–72 (7th Cir. 2008) (holding that in allowing its online customers to post housing ads that state racial preferences, Craigslist does not violate fair housing discrimination statutes).

  116. 116. Lending Club exacts a promise from its users not to discriminate. See Kadhim Shubber, Lenders Can’t Discriminate, But What About Investors?, Fin. Times: Alphaville (Jan. 13, 2016, 6:02 PM), http://ftalphaville.ft.com/2016/01/13/2150093/lenders-cant-discriminate-but-what-about-investors. Airbnb will require a similar commitment as of November 1, 2016. See infra note 119.

  117. [117]. See Kevin Montgomery, Google’s Ad System Has Become Too Big to Control, Wired (Aug. 8, 2015, 2:36 PM), http://www.wired.com/2015/07/googles-ad-system-become-big-control. Constraining the use of stereotyping of discriminatory information will be more difficult, however, and not always in the interests of the firm. Id.

  118. [118]. See Laura W. Murphy, Airbnb’s Work to Fight Discrimination and Build Inclusion: A Report Submitted to Airbnb (2016), http://blog.airbnb.com/wp-content/uploads/2016/09/
    REPORT_Airbnbs-Work-to-Fight-Discrimination-and-Build-Inclusion.pdf. Notably, the set of reforms does not appear to change Airbnb’s policy of requiring users to waive their right to bring a class action against the company as a condition of using the site. See Katie Benner, Airbnb Vows to Fight Racism, but Its Users Can’t Sue to Prompt Fairness, N.Y. Times (June 19, 2016), http://
    www.nytimes.com/2016/06/20/technology/airbnb-vows-to-fight-racism-but-its-users-cant-sue-to-prompt-fairness.html. For criticisms by two economists that the recent reforms do not go far enough in terms of incentives and transparency, see Ray Fisman & Michael Luca, Airbnb Isn’t Doing Enough, Slate (Sept. 9, 2016), http://www.slate.com/articles/business/the_dismal_science/
    2016/09/airbnb_needs_to_do_a_lot_more_to_fight_discrimination.html.

  119. [119]. This phenomenon may help to explain the rush by businesses to take actions designed to show disapproval of North Carolina’s House Bill 2 which, among other things, prohibits cities and counties from passing laws protecting their citizens from discrimination based on sexual orientation and gender identity. See Yezmin Villarreal, Here’s All the Business N.C. Has Lost Because of Anti-LGBT Bill, Advocate (Apr. 13, 2016, 4:57 PM), http://www.advocate.com/politics/2016/4/13/heres-all-business-nc-has-lost-because-anti-lgbt-bill.

  120. [120]. See, e.g., Chaney v. Plainfield Healthcare Ctr., 612 F.3d 908 (7th Cir. 2010) (finding a nursing home had created a hostile environment for a black nurse when it excluded that nurse from the rooms of residents who objected to being treated by black nurses).

  121. [121]. See Nick Duffy, Accommodation Website Airbnb Removes Listing That Banned Gay Couples, PinkNews (Nov. 23, 2014, 11:52 AM), http://www.pinknews.co.uk/2014/11/23/accomodation
    -website-airbnb-removes-listing-that-banned-gay-couples (following outrage over a listing that banned couples from Airbnb website, Airbnb pulled the listing); see also Emaneulla Grinberg, When It Comes to Dating Sites, Race Matters, CNN (Jan. 13, 2016, 11:32 AM), http://www.cnn.
    com/2016/01/13/living/where-white-people-meet-feat (discussing how a billboard company pulled a billboard for WhereWhitePeopleMeet.Com after public protest).

  122. [122]. See Maggie Haberman & Nick Corasaniti, Donald Trump, in Feud with Fox News, Shuns Debate, N.Y. Times (Jan. 26, 2016), http://www.nytimes.com/2016/01/27/us/politics/trump-feud-fox-debate.html.

  123. [123]. This example illustrates how the norm has changed over time. In a well-publicized 1983 case, a local television station anchorwoman, Christine Craft, was moved to an off-camera position because she apparently lost her good looks as she aged. The network successfully defended a Title VII case, in part, because of focus group research showing that Craft’s appearance had an adverse impact on her acceptance among viewers. See generally Craft v. Metromedia, Inc., 572 F. Supp. 868 (W.D. Mo. 1983), rev’d in part, 766 F.2d 1205 (8th Cir. 1985). For a first-hand account, see generally Christine Craft, Too Old, Too Ugly, and Not Deferential to Men (1988).

  124. [124]. See Washington Redskins Fan Goes Full Red-Skinned, TMZ Sports (Sept. 25, 2015, 6:51 AM), http://www.tmz.com/2015/09/25/washington-redskins-fan-full-redface-giants-game.

  125. [125]. See generally Ferrill v. The Parker Grp., Inc., 168 F.3d 468 (11th Cir. 1999).

  126. [126]. See generally Johnson v. Zema Sys. Corp., 170 F.3d 734 (7th Cir. 1999).

  127. [127]. See supra notes 8–9, 29 and accompanying text.

  128. [128]. See supra note 9 and accompanying text.

  129. [129]. For some suggestions, see Meera E. Deo, A Better Tenure Battle: Fighting Bias in Teaching Evaluations, 31 Colum. J. Gender & L. 7, 39–41 (2015).

  130. [130]. See Antonia Blumberg, Ultra-Orthodox Men Cause Flight Delay by Refusing to Sit by Women, Again, Huffington Post (Dec. 29, 2014, 3:03 PM), http://www.huffingtonpost.com/2014/
    12/29/orthodox-jewish-men-flight-delay_n_6391104.html; Isabel Kershner, She Was Asked to Switch Seats. Now She’s Charging El Al with Sexism, N.Y. Times (Feb. 26, 2016), http://www.nytimes.com/2016/02/27/world/middleeast/woman-81-to-sue-israeli-airline-over-seat-switch.html.

  131. [131]. On the general matter of why religious accommodations should not always trump civil rights claims, or be cost-free, see generally Mary Anne Case, Why “Live-and-Let-Live” Is Not a Viable Solution to the Difficult Problems of Religious Accommodation in the Age of Sexual Civil Rights, 88 S. Cal. L. Rev. 463 (2015).

  132. 133. For a description of these advantages, particularly with respect to discriminatory attitudes by employees, see Bartlett, supra note 24, at 1960–71.

  133. 134. See generally Estlund, supra note 24.

  134. 135. In the world of cricket, tickets for men’s and women’s games are also bundled. See Supriya Nair, Women’s Cricket Scores a Resounding Win in India, Fin. Times (Apr. 1, 2006), http://www.ft.com/cms/s/1eaa43b0-f75a-11e5-96db-fc683b5e52db.html.

  135. 136. See Part IV.A.

  136. 137. See generally Richard H. McAdams, The Expressive Powers of Law: Theories and Limits (2015) (describing the conditions that strengthen the expressive powers of law); Cass R. Sunstein, Social Norms and Social Roles, 96 Colum. L. Rev. 903 (1996) (exploring the role of law in changing and reinforcing social norms). Among the research, studies show injunctive normative information can suppress negative social norms. See generally Robert B. Cialdini, Crafting Normative Messages to Protect the Environment, 12 Current Direction in Psychol. Sci. 105 (2003); Robert B. Cialdini et al., Managing Social Norms for Persuasive Impact, 1 Soc. Influence 3 (2006). In the work discrimination context, see Grant & Sandberg, supra note 104.

  137. 138. Among all categories of civil litigation, discrimination cases are the hardest to win. Plaintiffs in discrimination cases do worse than defendants at every stage of litigation—settlements, pre-trial motions, trials, and appeals. See generally Kevin M. Clermont & Stewart J. Schwab, Employment Discrimination Plaintiffs in Federal Court: From Bad to Worse?, 3 Harv. L. & Pol’y Rev. 103 (2009); Kevin M. Clermont & Stewart J. Schwab, How Employment Discrimination Plaintiffs Fare in Federal Court, 1 J. Empirical Legal Stud. 429 (2004).

  138. 139. See generally J.E.B. v. Alabama ex rel. T.B., 511 U.S. 127 (1994) (holding that the state may not exercise its peremptory challenges on the basis of sex); Batson v. Kentucky, 476 U.S. 79 (1986) (holding that the state may not exercise its peremptory challenges based on race); SmithKline Beecham Corp. v. Abbott Labs., 740 F.3d 471 (9th Cir. 2014) (holding it unconstitutional for a party to use peremptory strikes on the basis of sexual orientation).

  139. 140. The exceptions involve blatant circumstances involving the complete elimination of prospective jurors based on race. See generally Foster v. Chatman, 136 S. Ct. 1737 (2016) (in a capital case, prosecutors highlighted each black prospective juror to indicate that the juror was black, ranked each black prospective juror against each other, and struck each black candidate from the jury pool).

  140. 141. As another example, some countries are considering bans on smoking in cars with children present. There is little chance of effectively enforcing such laws, but their advocates believe they would educate the public about the health risks of smoking and help change social norms about the acceptability of smoking around children. See Neil Thorpe, Smoking Ban in Cars with Children Difficult to Enforce, Conversation (Feb. 10, 2014, 12:53 AM), http://theconversation.com/smoking-ban-in-cars-with-children-difficult-to-enforce-22960.

*

Duke Law School.


**

Duke Law School. 


For comments, thanks to Joseph Blocher, Jamie Boyle, Al Brophy, Curt Bradley, Bernie Burk, Devon Carbado, Jonathan Cardi, Mary Anne Case, Guy-Uriel Charles, Lee Fennell, Joshua Fishman, Kim Krawiec, Maggie Lemos, David Lange, Ann Lipton, Richard McAdams, Darrell Miller, Louis Seidman, Patrick Shin, Naomi Schoenbaum, Neil Siegel, and Gregg Strauss, as well as to participants at workshops at Duke Law School, Indiana (Bloomington) Law School and the University of Chicago Law School, and at the Dean’s Lecture at Wake Forest School of Law. For research assistance, we thank Rose McKinley and Chaniece Mulligan, and for expert editing, we applaud Vol. 102 of the Iowa Law Review.