Structural inequality and the protectorate of discrimination law

Abstract

This article asks whether discrimination law should be symmetrical: whether it should offer the same level of protection to dominant and dominated groups. It articulates a structural inequality theory of the moral foundations of discrimination law and defends it against prominent alternatives, such as the view that discrimination is wrong because it is irrational or disrespectful. The paper then argues that while direct discrimination is symmetrical, indirect discrimination is asymmetrical. It cannot be claimed by those – men, or white persons – who are not at the sharp end of structural inequality. Furthermore, even dominated groups cannot claim to be indirectly discriminated against by just any law that has a disparate impact on them. If the structural inequality view is correct, the protectorate of discrimination law is not as extensive as is commonly assumed.

White persons, and men, sometimes claim to have been discriminated against. Our intuitions pull in different directions. We may think that disadvantage based on race or sex is often irrational and disrespectful, and wrongful on this basis. Or, we may think that anti-discriminatory norms are egalitarian in spirit and therefore not designed to protect members of dominant groups. In this paper, I adjudicate between these conflicting intuitions.

I do this by exploring the moral foundations of discrimination law. I hypothesise that what makes discrimination wrong is connected to the traits that it targets. Extant philosophical literature about discrimination has neglected the question of who the law should protect. Yet, this question – the question of the ‘protectorate’ of discrimination – is central to the philosophy of discrimination law. One task is to explain why the law focuses on the protection of certain traits and groups. Another is to ascertain whether all members of protected groups, including historically dominant members, such as white persons or men, should benefit from the protections of the law. Should discrimination law be symmetrical: should it offer identical levels of protection to members of dominant and dominated groups? In this paper, I defend what I call a structural inequality answer to these questions.

We need a working definition of discrimination to start with. Discrimination refers to the disadvantageous treatment of a P-person, compared to a non-P-person, because of P, where P is a certain trait or characteristic of the person. Discrimination can be direct or indirect. It is direct when P-persons are singled out by laws, rules or decisions. It is indirect when P-persons are affected by laws, rules or decisions without being singled out. Not all discrimination is wrongful, either at the bar of morality or of law. My aim in this paper is to explore the moral foundations of legal prohibitions of discrimination, both direct and indirect: the principled reasons why certain treatments of P-persons should be legally forbidden. My enquiry starts with an analysis of the implicit moral and political ideas that have underpinned the construction of the body of discrimination law. (The approach need not be deferential to all existing laws: instead, it asks, in a broadly Dworkinian interpretive spirit, what, on the best understanding of the law, the law ought to prohibit.) The most plausible theory of the salience of P, I argue, is that groups defined alongside P-axes, such as race, gender, sexuality and disability, are markers of what I call structural inequality. Discrimination law is best interpreted as a systemic response to structural inequality.

To begin thinking through the relevant issues, consider a green-eyed man. Can he be wrongfully discriminated against on grounds of (i) his eye colour or (ii) his sex? Eye colour is an idiosyncratic trait, in the sense that people with the same eye colour do not form a socially salient group. Maleness is an asymmetric trait, in the sense that, while sex is a protected characteristic, the protected social group is ordinarily thought to be women. To what extent should idiosyncratic and asymmetrical differential treatment be legally prohibited? Getting clear about both idiosyncratic and asymmetric discrimination will allow us to specify the protectorate of discrimination. On the structural inequality view, idiosyncratic discrimination is not (even presumptively) wrongful, and asymmetric discrimination is (presumptively) wrongful only when it is direct. While direct discrimination is symmetrical (it should protect dominant as well as dominated groups), indirect discrimination is structurally asymmetric (it should only protect dominated groups). Or so I shall argue. The paper is structured as follows. I first introduce and defend the Structural Inequality view, against a prominent alternative, the Idiosyncratic discrimination view. I explore the implications of the Structural Inequality view for the symmetry of direct discrimination. I then draw out the implications of the view for the asymmetry of indirect discrimination. If I am correct, there is no such thing as wrongful indirect discrimination against members of dominant groups. Men and white persons can complain of direct discrimination, but not of indirect discrimination. Towards the end of the paper, I further suggest that even dominated groups cannot claim to be indirectly discriminated against by just any law that has a disparate impact on them: the contested rule needs to track a previous practice of structural inequality. The upshot is that, on the Structural Inequality view, the protectorate of discrimination law is not as extensive as is commonly assumed.

Structural inequality v. idiosyncratic discrimination

It is conceptually impossible to discriminate without discriminating on the grounds of P. Plausibly, then, a key desideratum for any theory of wrongful discrimination – legal or moral – is to explain what makes differential treatment on the grounds of P wrongful. A popular view is that differential treatment is wrong because it attaches to morally arbitrary features of persons. There is no reason why the law should not prohibit differential treatment on the basis of any such feature: race, gender, but also eye colour, astrological sign, the first letter of one's name, etc. ¹ Call this idiosyncratic discrimination. It affects individuals as members of ad hoc groups: groups that have no social salience. On this theory, existing law focuses on membership in socially salient groups (call this Ps) but, in principle, wrongful discrimination could attach to any idiosyncratic trait or property of persons (call this Pi). In what follows, I consider two versions of the theory – the irrationality view and the disrespect view – and I show that both fail to articulate what is morally distinctive about disadvantageous treatment ‘because of P’ compared to other forms of irrational or disrespectful treatment. To that extent, theories of idiosyncratic discrimination fail to meet a core desideratum for a normative theory of discrimination law.

The irrationality view of discrimination locates the wrongfulness of discrimination in the fact that differential treatment is grounded in a property P that is irrelevant to the task or position for which the individual is considered. To deny someone a job because of the colour of their eyes or their astrological sign is as irrational as denying someone a position because of the colour of their skin, or their sex – when none of these properties is relevant to the successful completion of the job's requirements. To be sure, in some cases, the possession of some P is an occupational requirement: only people with good eyesight can be pilots, only men can be Catholic priests, and so forth. Yet, barring those tailored exemptions, what makes differential treatment on the basis of P wrongful is that the singling out of P lacks a rational basis.

The problem with the irrationality view is that it equates rationality with morality. Intuitively, discrimination can be wrongful, even when it is rational. Employers might correctly judge that skin colour is relevant when their customers would prefer not to be served by members of certain racial groups, or that sex is relevant when they calculate the pregnancy-related costs of hiring women. Irrelevance, therefore, cannot be the chief wrong-making feature of discrimination. Nor is it clear that irrationally singling out any P should be seen as distinctively wrongful, when compared to other irrational – arbitrary, unfair, whimsical – decisions that do not single out P. To see this, imagine a situation where the most suitable applicant for a job – call him Daniel – is not hired. Consider three scenarios.

Skin Colour. An employer dislikes people who, like Daniel, are black, and she denies him a job on this ground.

Eye Colour. An employer dislikes people who, like Daniel, have green eyes, and she denies him a job on this ground.

Throw in Bin. An employer acts on whim and throws Daniel's (and only Daniel's) application in the bin without looking at it.

Plainly, Throw in Bin is as irrational, arbitrary and unfair as Skin Colour and Eye Colour. The irrationality view does not have the resources to explain what exactly is problematic about treatment ‘because of P’.

The disrespect view, for its part, locates the wrong of differential treatment ‘because of P’ in a distinctive moral failure: a failure to respect persons qua persons. Benjamin Eidelson has provided the most sophisticated version of the respect-based theory of discrimination. On his view, we disrespect others when we fail to give proper deliberative consideration to their personhood either by failing to recognise them as persons of equal value or by failing to recognise them as autonomous individuals. When we do this by appeal to a trait or property P that distinguishes between different classes of people, we manifest ‘discriminatory disrespect’. For Eidelson, it makes no difference which feature of persons one is reacting to: it could be the size of their ear lobes, the colour of their eyes or the colour of their skin (Eidelson, 2015: 167). Discriminatory disrespect, therefore, is not limited to socially salient characteristics, such as Ps: discriminatory disrespect can attach to idiosyncratic features Pi.

A key question for Eidelson is whether discriminatory disrespect – disrespect on the basis of P – is a distinctive form of disrespect. What is distinctive about Eye Colour, compared to Throw in Bin? In both cases, Daniel is disrespected as an individual. In Eye Colour, the employer singles out the colour of his eyes at the expense of any other trait and information that makes him the individual that he is. In Throw in Bin, the employer does not single out anything about Daniel and therefore fails to take seriously any relevant feature of Daniel. In both cases, the employer disregards the person that he is, the specific features, choices and attributes that make him an applicant for that job. She fails to recognise him as an individual: she fails to respond appropriately to the morally relevant features that he exhibits. On the disrespect view, there seems to be no morally significant difference between Eye Colour and Throw in Bin. The disrespect view by Eidelson's own admission (Eidelson, 2015: 91) cannot explain what makes differential treatment ‘because of P’ distinctively wrongful.

Both the irrationality view and the disrespect view struggle to explain the moral difference between Eye Colour and Throw in Bin. There is a ready explanation for this: the difference is in fact miniscule. The harm of idiosyncratic differential treatment is comparable to that caused by sundry arbitrary, unfair and whimsical decisions. By contrast, the difference between Eye Colour and Skin Colour is morally salient. Plainly, Skin Colour is more harmful than Eye Colour because of its contribution to the perpetuation of the structural inequality that subordinates and subjugates black people. In Eye Colour, Daniel is only subjected to the capricious and idiosyncratic preferences of his whimsical would-be employer. His general prospects are not significantly affected by this one-time setback to his interests. The decision does not compound an existing structural disadvantage suffered by green-eyed people. It does not communicate to a wider society that it is acceptable to disadvantage green-eyed people. It does not lower the overall status of green-eyed people and does not contribute to their stigmatisation. In Skin Colour, by contrast, these graver social harms obtain (Lippert-Rasmussen, 2006: 169; 2014: 30–35). The upshot is that it is not primarily because racial categorization is disrespectful that racial discrimination is wrongful: it is wrongful because it maintains structural inequality.

Eidelson suggests that Eye Colour is as disrespectful as Skin Colour. Yet, the reason why this is the case also explains why Throw in Bin is disrespectful. On his view, the three cases disregard individuality and autonomy in similar ways. As a result, it is not clear what difference the presence of P in Eye Colour makes, as Eye Colour has identical wrong-making characteristics as Throw in Bin. This means that the disrespect account has no resources to account for the chief wrongful feature of discrimination: that it disadvantages people on the basis of P. Discrimination picks out some dimension of difference, and it is this singling out that is distinctively problematic. A whole gamut of decisions is irrational and arbitrary, and some may as well disrespect individuals. Yet, wrongful discrimination should not be equated with irrational and arbitrary decisions, lest we miss the specific wrong that discrimination law targets (Hellman, 2011: Ch. 5).

An alternative plausible view is that discrimination law is designed to tackle a specific evil: structural inequality. Structural inequality obtains when disadvantages in a number of significant dimensions – education, income, capital, power, recognition and status – cluster alongside social axes of distinction, such as race, gender, sexuality and disability. Such axes of distinction have typically been consolidated through long-term processes of historical domination, whereby some groups have accumulated and maintained greater power, resources, income, recognition and status compared to others. In contemporary societies ruled by ideals of equal rights, formal structures of domination have been largely dismantled. Yet, ‘enduring injustice’ – structural inequality – remains in the form both of institutionally reproduced exclusions and lingering prejudices and stereotypes (Nuti, 2019; Spinner-Halev, 2012; Young, 1990).

The key aim of discrimination law, I suggest, is to reduce the resulting gaps between historically dominant and dominated groups (Bagenstos, 2003; Khaitan, 2015; Moreau, 2020: Ch. 2; Shin, 2013). Structural inequality is of special concern for advocates both of distributive and relational equality: it signals an unjust distribution of opportunities and advantage and the enduring presence of problematic social attitudes. In addition to unfairly restricting opportunities and life chances, structural inequality gives rise to self-confirming stereotypes, it stigmatises certain people as social inferiors; it makes it less likely that people will relate as equals; and it renders people vulnerable to further domination. This account of structural inequality generates a plausible explanation of the function of prohibited grounds of discrimination. The traits commonly recognised in the law – race, gender, sexual orientation and disability – mark out the axes alongside which structural inequality has historically coalesced: via a history of group domination. Discrimination law protects what I call ‘structural groups’ – themselves made up of dominated groups (racial minorities, gender minorities, LGBTQ, the disabled) and their cognate groups (historically dominant groups, such as racial, gender or sexual majorities). ²

Discrimination law responds to the harms of structural inequality by promoting the integration of workplaces and other important areas of social life and by disrupting the pattern-sensitive social, cultural and economic processes that affect it. It tackles both its material and expressive dimensions. The structural inequality view is an effect-based view, which zooms on the broad social harms associated with the disadvantageous treatment of specific groups. Among such harms are expressive harms: plainly, discrimination on the basis of certain attitudes (racism, sexism) is particularly pernicious; in light of the expressive role, such attitudes play in the maintenance of patterns of structural inequality. On my view, an act of discrimination is (presumptively ³ ) wrong when:

Main Clause. It contributes to structural inequality, or

Secondary Clause. It expresses biased attitudes that legitimize structural inequality.

To be sure, the structural equality theory does not entirely discount the disrespect involved in discrimination. Among the sundry social harms associated with Ps-discrimination, one of the most grievous is the way it stigmatises and demeans its victims, against a background of structural inequality. Of our three scenarios, only Skin Colour evinces the kind of disrespect that is at the core of legal concerns about discrimination. Racial discrimination is different from Eye Colour: it is not, in fact, discrimination-on-ground-of-skin-colour. Rather, racial discrimination is wrong because it appeals to a social category – race – that is conventionally associated with a history of domination. Discrimination is not wrongful in virtue of being disrespectful, even if a particular kind of disrespect – demeaning disrespect – is one of the social harms of structural inequality.

The structural inequality account then offers a compelling account of the protectorate of discrimination because the most plausible alternative theory to it is philosophically over-inclusive. The idiosyncratic account would grant no special moral significance to differential treatment in relation to P, compared to any other arbitrary, unfair treatment.

Symmetric direct discrimination

I have argued that green-eyed persons cannot be wrongfully discriminated against because the irrationality or disrespect evinced in idiosyncratic treatment is not the type of harm that discrimination law concerns itself with. I now turn to the question of asymmetric treatment: whether white men, qua white and qua men, can be wrongfully discriminated against. Here is a variation on Skin Colour.

Skin colour (white). The employer dislikes people who, like Daniel, are white, and she denies him the job on this ground.

On the structural inequality view I have sketched, it would seem that white Daniel cannot be the victim of wrongful discrimination. If the point of discrimination law is to reduce structural inequality, then it is not designed to protect members of dominant groups. White Daniel, like green-eyed Daniel, is the victim of an irrational and arbitrary decision, but it is not clear, on the egalitarian view, that he suffers wrongful discrimination. Or so it would seem.

Strangely, the existing philosophical literature on discrimination has not had much to say about the issue of symmetric discrimination. ⁴ To be sure, those authors who defend respect-based theories (of the kind sceptically discussed in the previous section) have no problem accounting for it. On Eidelson's view, discrimination against white men is disrespectful simply because it fails to treat them as individuals. Yet we saw that this wrong – which also affects green-eyed persons, as well as those subjected to random, arbitrary treatment – cannot explain the specific wrong of singling out P. Turning now to effect-based theories, particularly egalitarian ones, a brief survey of the literature reveals a diversity of views.

Kasper Lippert-Rasmussen concedes that there is something wrong with what he calls ‘reverse racism’, but he does not specify what the wrong is (Lippert-Rasmussen, 2014: 139). On his harm-based prioritarian theory, it is difficult to see why white people could be protected by discrimination law, given that they are not likely – qua white people – to be among the worst-off and are not likely to be substantially harmed by acts of discrimination. In Deborah Hellman's expressivist egalitarian theory, discrimination is wrongful only when it expresses objectively demeaning attitudes. As members of dominant groups are not demeaned or stigmatised when they are differentially treated, the ethics of discrimination law is fundamentally asymmetric. The only reason why the law treats both groups symmetrically, on Hellman's account, is administrative convenience (Hellman, 2011: 37). Tarun Khaitan has proposed a congenial, more principled justification, which points out the deleterious expressive effect of asymmetry. As he puts it, ‘discriminating against members of a dominant group may have expressive repercussions which may also worsen the situation of the disadvantaged group’ (Khaitan, 2015: 231). The law's failure to protect dominant groups may feed backlash feelings of grievance and resentment, which themselves exacerbate identity competition and minority disadvantage (Siegel, 2011).

My reservation about Khaitan's all-things-considered consequentialist test is that it does not sufficiently distinguish between different kinds of expressive messages – those that directly work to legitimise structural inequality and those that might indirectly bring about undesirable side effects. As a result, it risks too prematurely casting suspicion on all race-conscious policies, including those designed to reduce structural inequality (such as affirmative action). Like Khaitan, I focus on the expressive effect of discriminatory acts and anti-discriminatory policies: the message such acts and policies convey, over and above their tangible effect (Anderson and Pildes, 2000; Hellman, 2000). Unlike him, however, I isolate only one such effect: the role played by certain attitudes – which I call biased attitudes – in the legitimization of structural inequality. In what follows, I explain how the secondary clause of my theory offers some protection for dominant groups without, however, casting premature suspicion on all race-conscious policies.

Structural inequality is typically underpinned by justifying ideologies. Gender and racial domination, for example, have been legitimised by the ideologies of racism and sexism. Such ideologies erect natural traits into socially visible ones and make them powerful markers of hierarchical social organisation. These ideologies have entrenched the unwarranted social salience of certain traits. They have legitimised prejudices, stereotypes and aversions (call them biases for short) alongside axes of distinction that should be irrelevant for most purposes of social organisation. The aim of discrimination law is not only to prohibit differential treatment that perpetuates structural inequality, but also to combat the ideological biases that feed and justify them. The secondary clause of my theory then allows that the attitudes that justify and rationalise the disadvantageous treatment have an identifiable wrong-making feature, yet one that is derivative of the main egalitarian clause.

Secondary clause. Wrongful discrimination expresses biased attitudes that legitimize structural inequality.

The expression of racial hostility and prejudice in societies, where the salience of race tracks a history of racial domination, helps maintain ideologies of inferiority and stigma. This expressive wrong is an additional wrong to the main wrong of contributing to structural inequality. Note that the wrong should not be located in the private intentions and mental states – conscious or unconscious – of discriminators ⁵ ; rather, it pertains to the social message communicated by the discriminatory action (Hellman, 2011). As with hate speech, what is wrong with the message that acts of wrongful discrimination send is the way in which they consolidate certain traits as markers of inferior social status (Waldron, 2012: 56–57). In expressively condemning such biased attitudes, discrimination law acts as a mode of ‘counter-speech’.

From the analysis so far, it might seem to follow that the secondary clause only bars prejudice and bias directed at dominated groups. This, however, misunderstands how ideological domination works. Norms of domination shape social expectations for dominant, as well as dominated groups. Consider, for example, male domination. It is underscored not only by the persistent subordination of women, but, more broadly, by the pervasiveness of sex-role differentiation along gender lines. Feminist lawyers, such as Supreme Court Justice Ruth Bader Ginsburg, well understood that the subordination of women is tied to masculine gender norms (Franklin, 2010; Hosein, 2015). Gender underpins norms of both femininity and masculinity, which together legitimise the system of domination we call patriarchy. This is why discriminating against a man on grounds of his sex is not the same, morally speaking, as discriminating against someone on grounds of the colour of his eyes. Differential treatment that is motivated by animus against men, or – more subtly – that communicates patriarchal assumptions about sex-role differentiation (e.g. the prejudice that males should not be nurses or stay-at-home parents) is wrongfully discriminatory. This is the case of prejudicial judgements on the basis of gender, where gender is a socially salient category that defines and consolidates sex-based roles and stereotypes alongside a socially irrelevant axis. ⁶ Biased judgements about men consolidate patriarchy and therefore indirectly contribute to structural inequality.

Racial domination works in a broadly similar way. In racially conscious societies, skin colour does not function as a merely idiosyncratic trait. Racial domination is underscored not only by persistent attitudes of stereotype and bias against racial minorities, but also by prejudice and hostility towards dominant racial groups. Anti-white racism is often a by-product of racial oppression (a mode of reverse racism) and, insofar as its public expression perpetuates racially biased attitudes and motivates differential treatment of whites qua whites, it is a legitimate target of discrimination law. Consider a recent Canadian case of employment discrimination against white employees (Eva and Others v. Spruce Hill Resort and Another. 2018 BCHRT 238). ⁷ The employer expressed a preference for Chinese workers on the grounds that white employees were less hard working and more likely to demand overtime pay or statutory holidays. The differential treatment disadvantaged the white workers while communicating racialized stereotypes that themselves contribute to maintaining racial hierarchies. My secondary clause condemns this treatment as wrongfully discriminatory. This is not because white workers are demeaned by the expression of anti-white prejudice: in societies where they are globally advantaged, they cannot be so demeaned. It is not either because failing to protect white persons might trigger resentment on their part. It is rather because anti-white prejudice contributes, however indirectly, to sustaining the racialized system of domination. ⁸ Biased attitudes against dominant groups in societies where the latter enjoy considerable power and privilege will form a small set. However, it is not an insignificant set, and it should be a legitimate target of discrimination law. ⁹

The secondary clause of my theory only targets those acts of differential treatment that also express biased attitudes. Biased attitudes in turn are wrong at the bar of legal morality not because they offend or disrespect, but because they sustain the very ideological system that supports racial domination. It should now be clear how my approach differs from Khaitan's all-things-considered consequentialism. I locate the relevant expressive effects of the symmetry of discrimination law exclusively in the communication of biased attitudes, not in racial classification itself. Race-conscious policies intended to reduce structural inequality, such as policies of affirmative action, are not presumptively wrongfully discriminatory as they do not express biased attitudes. This does not mean that the broader effects that Khaitan points to must not be taken into account when the best policies are to be designed. What it means is that undesirable social effects, such as majority resentment, should be considered alongside other possible deleterious consequences of affirmative action, such as the denial of applicants’ legitimate expectations, negative stereotypes about the ability of beneficiaries, the risks of balkanisation and social cohesion and so forth (Suk, 2013). While they all have to be factored in the all-things-considered assessment of the desirability of policies, they should be discounted when we articulate the normative foundations of anti-discrimination – lest the theory collapses into an ad hoc consequentialism. No doubt affirmative action policies are sometimes ill-advised but, crucially, this is not because they wrongfully discriminate.

This two-pronged theory of wrongful direct discrimination generates intuitive results for a number of familiar puzzles as follows.

Disadvantageous racist/sexist treatment of members of dominant groups. Hostility on the ground of sex and race is of a different kind from hostility on the basis of eye colour. It is likely to express biased attitudes, which maintain the ideological underpinnings of structural inequality.

Affirmative action. Race- and sex-conscious policies of preferential treatment for members of dominated groups are not presumptively wrongful. They can contribute to reducing structural inequality, and they are not grounded in biased mental states and attitudes. Affirmative action in favour of African-Americans in the US, for example, is not an expression of anti-white racism.

Rational and statistical discrimination. Many discretionary decisions cannot be criticized as expression of bias and can be rational responses to facts about our society. It is not an expression of prejudice if employers assume that women are more likely than men to take parental leave, and, therefore, that hiring them comes with certain costs. It may also make business sense for them to cater to their customers’ racist preferences. It might be statistically efficient for the police to engage in ‘racial profiling’ in certain contexts. Yet, on my theory, rational discrimination against dominated groups is presumptively wrongful merely by virtue of its broader effects: insofar as it has a predictably negative impact on larger social patterns of resources, opportunities and status. The fact that such forms of direct discrimination are underpinned by unbiased statistical evidence, rather than hostility or prejudice, does not impugn their wrongful nature.

The structural inequality theory of wrongful discrimination also delivers the right conclusion in the case I now turn to: indirect discrimination.

Asymmetric indirect discrimination

I have argued that members of dominant groups, such as men and white persons, can complain of wrongful direct discrimination. They can do so when they are differentially treated and disadvantaged out of biased attitudes, whose expressive effect is to legitimise the unwarranted social salience of race and gender. But, can they complain of wrongful indirect discrimination?

Recall that discrimination is direct when P-persons are singled out on the basis of P. It is indirect when P-persons are disadvantaged in relation to non-P-persons, without being singled out, by a provision, criterion or practice (PCP for short). A PCP can be legally wrongful when it is ‘fair in form, but discriminatory in operation’, as the US Supreme Court stated in its 1971 Griggs v Duke Power decision. In this landmark case, an employment qualification disadvantaging African-Americans was found to constitute unjustified disparate impact in light of the grossly inferior education provided to them in the segregated schools of North Carolina (Griggs v. Duke Power Co., 401 U.S. 424, 431 [1971]). ¹⁰ Indirect discrimination is usually found to be wrongful when it fails to meet some standard of ‘objective justification’ or means/aims proportionality. (In Griggs, there was insufficient business necessity for that particular employment qualification). Even before we get to the proportionality test, however, I argue that prior enquiries about the protectorate and scope of wrongful indirect discrimination are necessary. In line with the structural inequality theory, I will suggest that indirect discrimination is (presumptively) wrongful only when (i) it contributes to structural inequality – a forward-looking condition (this section), and (ii) it tracks structural inequality – a backward-looking condition (section ‘Tracking structural inequality’).

Recall that in cases of direct discrimination, the main clause of my theory – the act's contribution to structural inequality – is sufficient, but not necessary. The secondary clause, in turn, can be sufficient to condemn the expression of biased attitudes that indirectly contribute to the persistence of structural inequality, even when they target dominant groups. When it comes to the diagnostic of wrongful indirect discrimination, by contrast, the main clause is necessary and sufficient, and the secondary clause inert. This is because the prohibition of indirect discrimination is designed to root out practices that maintain structural inequality without expressing any bias or objectionable ideology. In general, no biased attitude is found in indirect discrimination. Therefore, indirect discrimination is wrongful when it perpetuates structural inequality, simpliciter. PCPs that incidentally disadvantage members of dominant groups are not wrongful: they might constitute an episodic, local disadvantage, but not a structural one.

On this view, indirect discrimination is structurally asymmetric. It should not be thought of as protecting every member of a suspect class in discrimination law: only members of dominated groups. To capture this in a slogan: indirect discrimination is wrong only if it tracks domination. In Griggs, disparate impact was a concern because African-Americans have been dominated in the US society: the educational disadvantage that they suffer is the product of a history of brutal domination marked by systemic slavery and continued segregation.

Bizarrely, the asymmetric structure of indirect discrimination has found few supporters in the literature on the philosophical foundations of discrimination law. There are two broad views. The first is the respect-based view. Here, the wrong of discrimination is exclusively located in the mental states that motivate and justify it. Insofar as indirect discrimination does not show anyone any disrespect, it does not count as wrongful discrimination at all (Eidelson, 2015; Gardner, 1998: 167, 182). This view puts into question the normative continuity between direct and indirect discrimination and marginalises indirect indiscrimination. This is an unfortunate result, given that most problematic exclusions today stem from the insidious workings of indirect discrimination, rather than the expressively transparent exclusions of direct discrimination.

The second family of views are effect-based theories. Because my theory places structural inequality as its central concern, some asymmetrical treatment of dominant and dominated groups is built into it. By contrast, theories that ground anti-discrimination in other values are more open to the possibility that dominant groups can be discriminated against, even indirectly. The challenge for them is to account for the distinctive harm suffered by members of dominant groups when they are episodically disadvantaged. Consider, for example, Sophia Moreau's suggestion that one sufficient condition for discrimination to be presumptively wrongful is that it infringes on what she calls ‘deliberative freedom’ (Moreau, 2010). This is the freedom to deliberate about our decisions and actions free from the assumptions that others make about our identity or the costs that come with our particular traits. On Moreau's view, members of both dominant and dominated groups can be wrongfully denied deliberative freedom, and they can be so both through direct and indirect discrimination.

The problem with Moreau's symmetrical proposal is that, as she recognises, we do not have a general right to deliberative freedom – there is no entitlement to be protected from the general costs associated with our exposure to sundry social interactions. Discussing a case similar to Eye Colour, Moreau concedes that the refusal to hire a green-eyed person might affect his deliberative freedom (he is not free to discard the colour of his eyes when applying to that position). However, Moreau concedes that this does not violate his right to deliberative freedom, since ‘[n]o one has a right not to have their green eyes held against them .. because this is not the kind of thing that fails to show respect for someone as a being capable of autonomy’ (Moreau, 2020: 115). But, then she needs to say more about which violations of deliberative freedom threaten autonomy in the relevant sense.

On my view, they do so, not when individuals are episodically or incidentally asked to bear some costs associated with some traits of theirs, but when those costs accumulate across different domains of social life. The right to deliberative freedom is violated when individuals are having constantly, rather than episodically, to live with their race or gender ‘before their eyes’ (in Moreau's own expression) – when they cannot afford not to factor others’ assumptions about these traits in many of the decisions they make about their lives (Calhoun, 2022; Moreau, 2020: 86). Members of groups that suffer structural inequality are those who have to endure repeated exposure to the negative assumptions of others. Violations of deliberative freedom, in sum, are only wrongful when they magnify structural inequality: Moreau's account collapses into the structural inequality account. In my view, mere differential impact on members of dominant groups neither perpetuates structural inequality nor validates the biased attitudes that legitimise it. As far as I can tell, only one prominent philosopher of discrimination endorses the asymmetry view. Lippert-Rasmussen explicitly restricts the protectorate of indirect discrimination to groups that have suffered past direct discrimination (Lippert-Rasmussen, 2014: 70). This suggestion seems to me essentially correct.

How can we justify it? We start by noting that there is a connection between past direct discrimination and the persistence of structural inequality. Over time, the wrongful harms of direct discrimination have ramified via PCPs that do not express racism, sexism or other biases and would be innocent, but for the connection between them and acts of past direct discrimination. These PCPs work to perpetuate inequality, not via facially differential treatment of disadvantaged groups, but through less visible, more structural and institutional mechanisms. For example, employment tests that privilege formal educational qualifications contribute to the continued exclusion of ethnic minority workers in contexts where – because of historical domination – they were denied valuable education and jobs. Workplace practices that burden those who are pregnant and give birth to children and who are disproportionately engaged in childcare and domestic labour contribute to perpetuate the historic marginalisation of women in the labour market. Such facially group-blind PCPs act as ‘built-in headwinds’ – to use the Griggs phrase – slowing down the fair integration of minorities. The institutional policies and practices of societies with a history of domination perpetuate the continued subordination of particular social groups without expressing any objectionable ideology. They – more than practices of direct discrimination – are now the chief cause of the stubborn persistence of structural inequality.

Such PCPs, by their very nature, only disadvantage historically dominated groups. As the norms and needs of dominant groups have permeated all spheres of social life, they are globally advantaged by existing institutions and practices. It is difficult to see how a question of indirect discrimination against them could even arise. Insofar as indirect discrimination points to a structural, systemic problem, then it can – logically – only affect minority, dominated groups. As I have argued throughout, the mere presence of a socially salient trait, or what the law calls a protected characteristic, does not suffice to trigger a presumption of wrongful discrimination. The fact that men or white persons are disadvantaged by some local provision might be unfair and regrettable, but need not constitute the specific wrong of discrimination. No relevant harm is done when a practice has a merely statistically disparate impact on members of a group without either communicating biased attitudes or perpetuating the broader pattern of inequality that discrimination law is designed to combat.

At this point, a critic might challenge the premise of my argument. Is it true that suspect PCPs do not communicate biased attitudes? Surely, indirectly discriminatory rules often contribute to the perpetuation of negative stereotypes about dominated groups. An apparently neutral employment test that is disproportionately failed by ethnic minority candidates will likely confirm prejudices about these candidates’ capacities. A workplace practice of holding meetings between 4 and 6 pm – making it difficult for those with childcare responsibilities to attend – will likely perpetuate the stereotype of women as more focused on domestic life than on their professional career (Moreau, 2020: 65–66). This is correct, and such effects are covered by the main clause of my theory. As I argued above, structural inequality is bad not only for distributive, but also for relational reasons, such as the fact that it gives rise to self-confirming stereotypes and feeds stigmatising social attitudes. Indirectly discriminatory rules often causally contribute to a state of affairs – structural inequality – that has undeniably deleterious expressive effects.

Yet, this is different from saying that the general rule itself has a distinctive social meaning. A wide range of workplace practices, employment tests, etc, are often said to be ‘neutral’ precisely because they are expressively silent. They do not express any sexist or racist bias – by contrast to those practices of direct discrimination that justify singling out some groups via the communication of objectionable ideologies. When no group is facially singled out by the contested rule, it is implausible to claim that an expressive message about them is being conveyed. This is why the secondary clause of my theory is inoperative in cases of indirect discrimination. The structural inequality theory urges us to direct our attention to the effects of expressively silent, generally applicable PCP on the actual perpetuation of structural inequality (including its expressive dimensions). As a result, it is best equipped to detect insidious forms of institutional and unconscious discrimination. Indirect discrimination is not wrongful (when it is wrongful) because of what it expresses, but because of its impact on structural inequality.

Tracking structural inequality

A crucial question remains. When exactly can a PCP be deemed to perpetuate structural inequality, such that it presumptively constitutes wrongful indirect discrimination? After all, sundry practices and policies tend disproportionately to affect historically dominated groups. Racial minorities suffer from policies that disadvantage the poor and the unemployed; religious people's life plans are routinely disrupted by secular laws; and women's access to professions requiring physical strength is limited. In these cases, members of protected groups are disadvantaged by general rules that do not single them out. But, are these ipso facto wrongful? The standard legal solution to this problem is to deploy a necessity or proportionality test: a PCP indirectly discriminates when it fails some standard of objective justification or means/aims proportionality.

Here, I want to raise a more principled scoping issue, whereby the set of suspect PCP needs to be narrowed before we apply a proportionality test. The issue is this. The set of PCPs that disproportionately disadvantage dominated groups is potentially extensive, making the charge of wrongful indirect discrimination all-pervasive, and therefore potentially vacuous. Note that this scoping problem does not arise in direct discrimination cases, where there is no uncertainty as to what the suspect set of PCPs is. In direct discrimination cases, classification (the facial targeting of a P-group) is the trigger for scrutiny because of its potential expressive effects. By contrast, the suspect set of PCP for purposes of indirect discrimination is indeterminate. Is mere impact sufficient, or should there be a connection between the suspect PCP and past discrimination? This crucial question has been somewhat neglected in extant literature. In what follows, I sketch two broad views – which I call the No-Connection and the Compounding views. My criticisms will form the basis for the defence of an alternative view, the Tracking view, which builds on their respective strengths without inheriting their weaknesses.

No connection

A PCP that indirectly discriminates against some groups is wrongful when it perpetuates structural inequality, regardless of whether the PCP is connected with past discrimination – hence ‘No-Connection’. On this view, indirect discrimination law serves the moral ends of distributive justice: it prohibits adding to the disadvantage of the already disadvantaged. A PCP that makes the worst-off even more worse-off is suspect, at the bar of the No-Connection view, even if the PCP bears no connection to the cause of their disadvantage.

Can the No Connection view account for the distinctive wrong of indirect discrimination? I will argue that it cannot – and that this counts against the view. Consider Lippert-Rasmussen's sophisticated theory. His version of the No Connection view is rooted in a theory of justice that identifies general duties of right and wrong actions derived from fair distributive principles. Lippert-Rasmussen favours a ‘desert prioritarian’ view, according to which, it is unjust to cause additional harm to people who are already unjustly worse off (Lippert-Rasmussen, 2014: 165–170). To what extent does the body of discrimination law meet this prioritarian test? Lippert-Rasmussen notes that the protectorate of discrimination law is made up of members of what he calls ‘salient groups’ defined alongside sex, race and so forth (Lippert-Rasmussen, 2014: 30). ¹¹ Indirect discrimination law protects only members of those salient groups that ‘suffer or have suffered direct discrimination’ (ie, it is asymmetric). Any claim of indirect discrimination must be ‘suitably related’ to instances of direct discrimination: it concerns ‘only disadvantage that perpetuates disadvantage resulting from past or present discrimination’ (Lippert-Rasmussen, 2014: 70–71).

What exactly is wrongful, however, about further disadvantaging members of those groups who suffered discrimination in the past? While the structural inequality theory gives a straightforward answer to this question – it points to the distinctive harms of group-based historical cumulative disadvantage – it is less clear that Lippert-Rasmussen's prioritarian theory does so. The problem is that the general category of the worst-off is not co-extensive with that protected by discrimination law. The prioritarian theory of justice, as a result, is vulnerable to an objection similar to the one I raised to Eidelson's theory. It cannot justify the protectorate of discrimination law: it cannot explain the moral relevance of P (the categories protected by discrimination law).

The mismatch between the groups protected by discrimination law and the unjustly worst-off (according to prioritarian justice) is observable on many levels. First, the groups making up the worst-off, on several dimensions, need not be P-groups. The worst-off overall might be people on low income; yet, discrimination law protects groups on grounds of race, sex and so forth. Second, even if the latter groups are, at an aggregate level, the most disadvantaged in relation to specific burdens, individual members of such groups are not necessarily the most disadvantaged, nor are they always disadvantaged by indirectly discriminatory PCP. For example, some individual women might be advantaged by a PCP that indirectly discriminates against women. In professions requiring height or strength, the minority of women who meet the criterion might be advantaged by the rule (in that they do not have to compete with many other women). Selection tests with racially disparate impact can be failed by members of racial minorities for reasons that are independent of their minority status. At the bar of prioritarian justice, the contested PCPs in these two cases do not unjustly disadvantage these individuals. Yet, they seem paradigmatic cases of unjust (group) indirect discrimination.

The third example of inadequacy of the prioritarian diagnosis of indirect discrimination is that Lippert-Rasmussen's theory of justice does not single out one cause of disadvantage over another. To be sure, the disadvantage has to be unjust – for example, on the desert prioritarian theory, it must have been brought about by circumstances external to the individual's own choices. Yet, it can be a case of bad luck and need not be connected to a history of structural inequality. On Lippert-Rassmussen's theory, a PCP that burdens a victim of prior discrimination is not morally worse than a PCP that burdens another unjustly worse-off person: what matters is whether additional marginal harms increase overall levels of harm. Reviewing cases such as these, Lippert-Rasmussen logically concludes that indirect discrimination is often not unjust at the bar of prioritarian justice (Lippert-Rasmussen, 2022a, 2022b).^. Differential disadvantage falling on salient groups is not worse than, or distinctive as opposed to, disadvantage falling on non-socially salient groups (Lippert-Rasmussen, 2014: 178, 2022a: 89–90).

This, on my interpretive approach to discrimination law, provides a reason to reject the No-Connection view. Because the view is grounded in a general moral theory, it struggles to articulate the specific normative concerns that underpin discrimination law: it does not adequately explain what makes differential treatment on the grounds of P distinctively wrongful. The only way to explain the distinct wrong of indirect discrimination is to suggest that it is a causal effect of past direct discrimination. Therefore, we need an alternative approach, which zooms more precisely onto the relevant mechanisms of discrimination.

Compounding

Consider classic cases, such as Griggs. In Griggs, there is a clear sense in which African-American employees were disadvantaged by the contested selection test because of their race. This is so not because racial motivations guided the employer's policy nor because the test communicated racial bias nor because every single African-American employee would fail the test. African-Americans were disadvantaged as a group in the sense that previous unjust race-based discrimination in educational opportunities explains why a seemingly group-blind test specifically disadvantages African-Americans. Griggs clarifies how indirect discrimination sheds light on ‘the reason why’: the causal mechanism that – however remotely and unintentionally – explains why certain groups are disadvantaged and continue to be so. The No Connection view, as we saw, does away with this causal connection: as a result, it cannot account for the specific wrong in Griggs.

An alternative account has been proposed by Deborah Hellman: the Compounding account. An agent compounds a prior injustice when the agent ‘amplifies the harm’ that the injustice inflicted and ‘takes the fact of that person's victimisation or its effects as her reason for acting’ (Hellman, 2018, 2020). Here, there is a tighter connection between the suspect PCP and past discrimination insofar as the former ‘compounds’ previous injustice. In Griggs, African-Americans were denied education, and then were disadvantaged on the job market because they lacked education. In Feeney 1979, the US Supreme Court challenged a policy that granted preferential employment to veterans in the civil service. Although the policy was facially neutral and had not been adopted in order to keep women out of the civil service, it in effect ‘permitted the state to compound the prior injustice of women's exclusion from military service’, in Hellman's view (2020: 481).

Compared to the No-Connection view, the Compounding view adds welcome determinacy to diagnoses of indirect discrimination. We now have a connection between suspect PCPs and prior discrimination. The Compounding view, however, is too restrictive. It contains two components, only one of which turns out to be necessary. The first, the Implication component, consists in the discriminator taking ‘the prior injustice or its effects’ as the reason for their action. The second, the Amplification component, consists in ‘making the harm caused by the prior injustice worse or causing it to lead to another harm in a new sphere of life’ (Lippert-Rasmussen, 2023). In what follows, I argue that Implication is too strong – leaving us with Amplification as the correct account of the wrong of indirect discrimination.

The notion of Implication assumes that the discriminator takes the fact that someone has suffered an injustice as a reason to perform an action that will further harm that person. This, however, gives too much weight to discriminators’ reasons. On the one hand, it is not clear that agents’ intentions are neatly predictive of the actual harm that their action produces (Lippert-Rasmussen, 2023). On the other hand, and more importantly for my purposes, the majority of suspect PCP cannot be traced back to reasons featuring in conscious deliberations of identifiable agents. ¹² Most cases of suspect PCP relate to general rules of social or workplace organisation that have not been put in place deliberately to exclude anyone. For example, agents who apply merit-based criteria do not orient their agency to the prior wrongs in any way. Even if they know about them, these wrongs often do not figure at all in the justifying reason for their action (Eidelson, 2021: 273). Perhaps we could suggest, as Moreau does, that failure to change or remove PCPs that have been shown incidentally to disadvantage dominated groups is a form of culpable negligence. ¹³ Yet, as Moreau herself insists, the wrong of indirect discrimination is best thought of as an institutional or structural wrong, not a personal one. On the view I favour, indirect discrimination points to a systemic, institutional failure, which primarily requires collective action (the elimination of certain tests, the reform of certain practices), rather than the correction of personal legal wrongs. ¹⁴

We are left, therefore, with the Amplification component of the Compounding View. Hellman's account requires not only that the groups affected have suffered injustice, but also that the policy at issue is somehow related to that injustice (Hellman, 2018: 121). This is on the right track. The difference with Lippert-Rasmussen's theory is that what matters is not the marginal amount of harm that is inflicted on previously excluded groups – on a consequentialist, all-things-considered calculus, this harm might well count for very little. Rather, a plausible aim of indirect discrimination law is to condemn the often invisible reproduction of the mechanisms that caused exclusion in the first place. Here Hellman's approach – when purged of its restrictive fault-based Implication component – joins hands with theories of structural or institutional racism/sexism (Attrey, 2021). Indirect discrimination law has an expressive function: it communicates that it is not acceptable to leave in place institutional arrangements that insidiously and often silently amplify prior discriminatory exclusions.

Therefore, I propose to revise Hellman's account. We should keep her focus on the connection between contested PCP and specific prior exclusions (exclusion from education, from military service, etc) while dispensing with the overly demanding second prong of her ‘compounding account’: the emphasis on personal implication. Here is my preferred view:

Tracking

A PCP is indirectly discriminatory if it tracks a particular historical exclusion (pace the No-Connection view) without necessarily entailing any agent's implication or complicity with injustice (pace the Compounding view). The notion of tracking exclusion is not as narrow as that of compounding injustice; yet, it captures the specific mechanisms through which exclusions in the past are maintained and perpetuated as injustices in the present.

On this view, there is a presumption that PCPs that incidentally put racial or sexual minorities at a disadvantage are suspect. After all, why should race or sex be correlated to disadvantage if it were not for a history of structural equality? Disparate impact on P-groups must be presumed to be rooted in causal mechanisms that reproduce or maintain previous injustices, even if the detailed causal mechanisms of reproduction of structural inequality are elusive. The presumption of wrongful indirection discrimination will be defeated only in the rare cases where the disadvantages are clearly shown to be causally independent of prior unjust discrimination against the group. One such case is hinted at by Hellman. She argues that (direct) discrimination on grounds of sex is often presumptively wrong, because it is most likely to be causally related to sex injustice. But in some cases it is not – when, for example, it benignly accommodates biological differences such as average differences in weight and strength (Hellman, 2020).

A PCP can track historical exclusion in two main ways. First, it can operate as a form of reiterated injustice. This happens when the application of a neutral (say, meritocratic) provision amplifies the harms caused by a separate, previous action that itself constituted an injustice, even if the specific causal connection between group membership and PCP-disadvantage is not demonstrable. ¹⁵ This was the scenario of Griggs, and the model for Hellman's Compounding account. A number of selection and employments tests, which amplify the harm of deliberate previous exclusions from education and training, can be said to constitute reiterated injustices.

Second, a PCP can function as a mode of continued exclusion. Consider what Moreau calls ‘structural accommodations’ (Moreau, 2020: 42–32, 56–63). These are policies, practices and physical structures that look neutral, yet tacitly function to privilege the interests and needs of dominant groups while continuing to marginalise historically excluded groups. As a result, the needs of dominant groups are perceived to be normal and natural, and those of dominated groups exceptional and odd. For example, physical spaces are constructed for normal human bodies and are not accessible to the disabled. Work uniforms are designed for ‘standardized bodies’ – not for the obese, the pregnant or those wearing religious dress. Continued exclusion often takes place through the reproduction of un-reflected upon cultural norms and assumptions – such as the norm, prevalent in Christian-dominated societies, that workers should be bare-headed and clean-shaven. Demands of accommodation by dominated groups – the disabled, women, religious minorities – are not demands for special, differentiated rights: they are, rather, egalitarian demands of integration on an equal basis. In these specific cases, accommodation is a response to a distinct wrong: the wrong of indirect discrimination.

The upshot of the Tracking view is two-fold. First, the view provides further support for my core thesis: there is no presumption of indirect discrimination against dominant groups. On the Tracking view, some connection – albeit not necessarily a tight one – is required between a PCP and a history of discrimination. Second, the view further suggests that there is no presumption of indirect discrimination against dominated groups in (the rare) cases when the PCP does not track any historical exclusion or injustice. The Tracking view offers a systematic method to reach coherent judgements in controversies about indirect discrimination, as I shall now suggest. Some of the following examples are loosely based on actual legal cases in the UK, US and Canada, although the description of the cases is simplified and stylised for illustrative purposes.

Academic Tenure Clock. In North American universities, academics apply for permanent employment roughly 5 years after initial appointment, pursuant to them demonstrating their scholarly credentials by publishing academic monographs and articles. As the 5-year ‘tenure clock’ typically runs when employees are in their 30s, it disproportionately burdens women of childbearing age. This is a case of reiterated injustice: the historical exclusion of women from universities in the past is compounded by them facing greater obstacles to full participation in the present.

Friday prayer. A Muslim teacher in a London primary school, Mr Ahmad, was fired for taking extra time off to pray on Fridays at lunch break. ¹⁶ This can be construed as a case of indirect discrimination, given background conditions. The Monday-to-Friday working week assumes Christian patterns of religious observance as the norm around which it designs expectations of work availability. By contrast to religious minorities, such as Muslims or Jews, Christians do not have to choose between weekly pray and full-time work. There might be disagreement as to whether Sunday laws constituted a historical injustice. Were such laws established to secure the hegemony of Christian faith and exclude minorities, or simply inadvertently, to suit the needs of the great majority of the population at the time? Either way - whether as a mode of reiterated injustice or continued exclusion – the relevant PCP (‘no time off on Friday’) tracks a troubling historical inequality and raises a (defeasible) presumption of indirect discrimination.

These are two clear cases of presumptively wrongful indirect discrimination in the Tracking mode. Now, consider two cases that do not presumptively qualify as wrongful discrimination because of the absence of the tracking condition.

Seniority rules for Muslim chaplains. A Muslim prison chaplain challenged the pay scale for prison chaplains on the grounds that it was indirectly discriminatory. Until recently, only Christian prison chaplains were employed, there being too few Muslim prisons. Because prison chaplains’ pay is based on years of service, no Muslim chaplain had joined their Christian counterparts at the top of the pay scale. Assuming it is true, as the Prison Service claimed, that there were not enough Muslim prisoners to justify employing chaplains on a salaried basis prior to 2002 - not itself an injustice or unjustified exclusion -, there is no presumption of wrongful indirect discrimination. ¹⁷

Sunday work. A care assistant in a children's home and a practicing Christian, Celestina Mba, was dismissed for refusing to work on a Sunday and took the London borough of Merton to court for indirect religious discrimination. ¹⁸ The provision that all employees of the care home be available every day does not track any history of privilege or advantage. Mba is not discriminated against in relation to Muslim, Jewish or Sabbatarian employees: all employees have to fit in with a religiously neutral working week. Nor is she discriminated against in relation to her non-religious colleagues: this would assume, implausibly, that any secular law burdening religious believers is (presumptively) wrongfully discriminatory. ¹⁹

Conclusion

In this paper, I have defended the structural inequality theory of the philosophical foundations of discrimination law. I have argued that discrimination law rightly focuses on structural groups because it is designed to combat the persistence of patterns of structural inequality. Members of dominant groups can be protected from the expression of biased attitudes that legitimise structural inequality. Yet, they are not to be protected from the mere incidental impact of general rules and laws. Only members of dominated groups can avail themselves of indirect discrimination protections. Yet, even they cannot presumptively challenge all burdensome rules: only those that track past exclusions or injustices.

On the structural inequality view, the protectorate of discrimination law is not as extensive as is usually assumed. By contrast to other theories, my theory of the grounds of discrimination is deliberately narrow. It restricts legal discrimination to a specific wrong, not to be conflated with other wrongs such as failures of individual respect, denial of fair opportunity, or infringement of religious freedom. In assessing the evolution of discrimination law – and the way in which sundry groups today claim to benefit from its protections – we would do well not to lose sight of its chief purpose: the rectification of patterns of structural inequality.