The Protected Grounds of Religion and Belief: Lessons for EU Non-Discrimination Law

Abstract The article draws lessons for EU non-discrimination law from the protected grounds of religion and belief through a discussion of the CJEU’s headscarf judgments. The article has two ambitions. First, the judgments are used to draw broader lessons for EU non-discrimination law, in relation to the distinction between and the justification of direct and indirect discrimination, as well as the purpose of protecting against (religious) discrimination. Second, these lessons are used to analyze the headscarf judgments and the criticism directed at them. While there is widespread agreement that the CJEU erred in these judgments, there is little agreement as to what mistakes were made. Through a discussion of these judgments, the article clarifies the difference between direct and indirect discrimination and the justification of both forms of discrimination. It is argued that the headscarf cases correctly distinguished between direct and indirect discrimination, and that the problem lies in the justificatory burden for indirect discriminatory measures, which was set too low by the CJEU.


A. Introduction
Compare the following two company policies. Company A prohibits Muslim employees from wearing the Islamic headscarf at work; incidentally, it also prohibits the manifestation of other religious, philosophical, or political beliefs. Company B allows Muslim employees to wear religious clothing except when visiting clients who do not permit the wearing of such clothing. Which of the two policies is more objectionable? There are credible reasons to think the more objectionable is Company A's policy. After all, for Muslims forced to choose between their religious beliefs and their jobs, it is cold comfort that non-Muslim religious colleagues working for company A may face a similar choice. In addition, company A's policy seems more reprehensible because of its more adverse consequences: It bans the Muslim headscarf at all times, in contrast with company B, which allows the wearing of the headscarf, except when this is not desired by clients.
The Court of Justice of the European Union (CJEU) reached the opposite conclusion in respect to such policies. Policy A is the defendant's policy in the Achbita case, while Policy B is the policy that was challenged in Bougnaoui. 1 Both cases concerned employees who were dismissed for refusing to remove their Islamic headscarves at work, in accordance with policies meant to meet the wishes of at the same time seems implicated by the law, may be the reason that it has been difficult to pin down exactly where, if at all, the CJEU has gone astray.
But these rulings, and the challenge they pose, also present an opportunity: They make it possible, even compel us, to reflect on non-discrimination law more broadly, beyond the specific factual circumstances of the cases, and even beyond the particular grounds of religion and belief. The second, more general ambition of this article, is to use the headscarf cases to draw broader lessons for EU non-discrimination law, in relation to the distinction between and justification of direct and indirect discrimination, as well as the more general purpose of protecting against (religious) discrimination. Why do we protect religion and belief from discrimination, alongside personal characteristics such as race, gender, and age? Wherein lies the evil of (religious) discrimination and what remedies does the law offer? Revisiting such questions is not only valuable in its own right, but also yields important insights into the headscarf judgments and the criticism that has been levelled at them. If we want to understand why the CJEU failed to recognize the wrong in religious discrimination in the headscarf cases, we need to reflect in greater detail on the purpose of non-discrimination law and the point of including religion and belief within the scope of its protection. Clarifying these questions is important not just in relation to future disputes on the right to wear religious clothing at work, but also given the fact that religious discrimination disputes seem to be coming more frequently before the CJEU. 6 This article is structured as follows. Section B clarifies the purpose of EU non-discrimination law, explaining that it seeks to reduce the disadvantages experienced by salient social groups. Both direct and indirect discrimination seek to remedy this same injustice, which the CJEU failed to recognize in the headscarf judgments. Section C defines the place of religion and belief in nondiscrimination law. An implicit assumption behind the headscarf judgments is that religion and belief deserve less protection than other protected grounds, such as race or disability. This assumption is shown to be incorrect: The rationale for the prohibition of religious discrimination is the same as the rationale for prohibiting discrimination based on other protected grounds. Having clarified the evil of religious discrimination, the article will proceed to discuss the two main points of criticism directed at the headscarf cases, to gain a fuller understanding of how EU non-discrimination law seeks to achieve its purpose. Section D clarifies where the boundary between direct and indirect discrimination lies and shows that the CJEU was in principle correct not to find direct discrimination in Achbita, IX v Wabe, and L.F. v. SCRL. Section E discusses the main justifications for direct and indirect discrimination and explains that these judgments suffered from the underlying assumption that direct discrimination involves a different and greater moral wrong than indirect discrimination. This explains why the CJEU set the justificatory burden for indirectly discriminatory measures too low. The problem is that in the context of indirect discrimination the CJEU accepted as legitimate an objective that it would never have accepted as legitimate in the context of direct discrimination.

B. The Purpose of EU Non-Discrimination Law
The outcomes of Bougnaoui and Achbita, and indirectly, therefore, of IX v. Wabe and L.F. v. SCRL, seems paradoxical: The more harmful measure was deemed more permissible because it did not constitute direct discrimination. Some may believe that the outcome, paradoxical as it may seem, was a logical consequence of the Framework Directive's application, which distinguishes between direct and indirect discrimination and provides a more open-ended justification for the latter. Defenders of the CJEU may argue that it simply applied this distinction. This section argues that 6 See, e.g., this view is incorrect and that the distinction between direct and indirect discrimination is not as straightforward: How it is drawn, and what justifications for either form of discrimination should be treated as legitimate, depends on deeper assumptions about the purpose of non-discrimination law. This section seeks to clarify its purpose, showing that indirect discrimination must not, by definition, be judged more benignly than direct discrimination.
Initially, the EU prohibited discrimination on grounds of nationality and gender as part of its ambition to create a level playing field for companies operating within the internal market. 7 The Treaty of Maastricht, however, expanded the EU's competence to combat discrimination, partly moving beyond this market rationale. The Framework Directive clearly pursues goals other than purely economic ones. It is "a specific expression . . . of the general prohibition of discrimination laid down in Article 21 of the Charter," 8 which, as former AG Sharpston put it, intends "to ensure that everyone can access the employment market under conditions that respect their identity and their dignity." 9 This, however, leaves us with an important question that this section addresses and that must be answered to better understand the headscarf judgments' shortcomings: What is the wrong that hinders access to the employment market that EU non-discrimination law seeks to remedy?

I. An Effects-Based View of EU Non-Discrimination Law
There are, broadly speaking, two distinct views on this question. According to the more traditional view, direct discrimination and indirect discrimination trace different wrongs, being each other's "logical counterparts." 10 The former is more objectionable because it involves culpable intent, 11 while the latter is more excusable, committed negligently or incidentally rather than intentionally. Although the CJEU did not spell out its position, the headscarf judgments seemingly rest on the more traditional understanding; they applied very different standards of justification to direct and indirect discrimination. On the alternative view defended here, direct discrimination does not always involve a greater moral wrong than indirect discrimination. Following established theories of nondiscrimination law, it is explained that EU non-discrimination law is centrally concerned with remedying the disadvantages burdening members of vulnerable social groups, and that the effects for the victim are more important than the motives of the perpetrator. This suggests that the headscarf cases hinge on an inflated notion of the distinction between direct and indirect discrimination.
Theories of non-discrimination law have appealed to different values in an attempt to articulate the moral wrong of discrimination-including human freedom, 12 human dignity, 13 personal autonomy, 14 and equality. 15 According to pluralist theories, moreover, the wrong of discrimination can be located in multiple values at once. 16 But despite disagreement about the value(s) on which the moral wrong of discrimination is based, these theories provide a remarkably coherent account of the aims and function of non-discrimination law. Group discrimination has emerged as the central concept in theories of non-discrimination law, 17 according to which its central purpose is to address discrimination against members of socially salient groups. 18 A plausible explanation for their convergence around the concept of group discrimination is that these theories have been influenced by, and been developed in conjunction with, legal practice. 19 As Khaitan explains, non-discrimination laws within different legal systems bear important resemblances: They seek to reduce the advantage gaps that exist between members of protected social groups. 20 A first characteristic these laws share is that they identify specific personal characteristics that they protect from discrimination, the so-called protected grounds. 21 Differentiation on the basis of personal characteristics that are not identified as protected grounds is not, in principle, prohibited discrimination. By way of illustration, EU non-discrimination law prohibits discrimination on the grounds of racial and ethnic origin. 22 Eye color is not among the protected personal characteristics and may therefore be relied on to differentiate between persons, unless a distinction based on eye color would amount to racial discrimination. 23 Why does the law single out and protect specific personal characteristics? First, the protected personal characteristics are capable of categorising persons into two or more groups. For example, religion can be categorised into Islam, Christianity, Hinduism, and so forth. A law that applies to all persons, such as a law that guarantees everyone a minimum level of subsistence, is not a norm of non-discrimination law for the reason that personhood is not a protected ground. 24 However, as the personal characteristic eye color tells us, not all characteristics that can be divided into more than one group are protected. Typically, only personal characteristics that seriously affect the allocation of opportunities and resources in our society are so, for example, characteristics that are shared by a socially salient group-a group whose membership is "important to the structure of social interactions across a wide range of social contexts." 25 As Shin explains, this is the case of groups "in respect of which, as a matter of empirical psychological fact, people have a tendency to harbour prejudicial, stereotypical, and bigoted beliefs or attitudes-both conscious and unconscious." 26 Such "inequality-laden attitudes" 27 make that certain groups-women, black people, religious minorities, persons with disabilities, suffer from significant political, sociocultural, and material disadvantages. 28 The fact that EU non-discrimination law protects personal characteristics that we associate with group-based disadvantage may seem to suggest that its ultimate aim is to reduce and eliminate socio-political and economic disadvantages between social groups. 29  features indeed support this view. For example, Article 7 of the Framework Directive authorises the adoption of positive measures to prevent or compensate for disadvantages linked to a protected characteristic. 30 The fundamental premise underlying this provision is that our society is organised and structured in a way that favors dominant majority groups, and that affirmative action is one means to strengthen the opportunities of disadvantaged groups. 31 This provision may be used, for example, to recognize religious diversity and grant specific benefits to religious minorities, for instance, days off to celebrate religious holidays, to reduce the disadvantages they experience compared to the Christian majority. 32 However, EU non-discrimination law is not exclusively concerned with addressing relative group disadvantage. It operates through a framework of individual rights; victims do not have to prove that the harm they suffer harms the social group to which they belong, or even that they belong to a disadvantaged social group. 33 As Waddington notes, the legal framework "is, in general symmetrical, meaning that it protects from discrimination both the majority or advantaged group and the minority or disadvantaged group." 34 Imagine a Muslim employer, in an overwhelmingly Christian society, refusing a job to a Christian applicant because of his antipathy towards Christians. 35 We think this amounts to discrimination, even if the disadvantaged person belongs to the dominant religious group and even if this group is not clearly burdened by the employer's intolerance. It certainly seems to amount to discrimination under EU law. For example, in Roca Álvarez, a male litigant could bring a claim for discrimination in respect of parental leave provisions, even though he could hardly claim to belong to a disadvantaged group. 36 In other words, EU non-discrimination law is as much about remedying individual wrongs as it is about group-based wrongs. 37 But this does not alter the fact that an effects-based view seems better equipped to account for essential features of the law, such as that it allows for positive measures to prevent or compensate for disadvantages related to a protected ground. It would also explain the fact that an intention to discriminate is not a necessary condition for establishing direct discrimination. 38 This effectsbased view corresponds to prevailing perceptions of non-discrimination law in almost all of the jurisdictions surveyed, where direct and indirect discrimination are increasingly treated as falling on a conceptual continuum rather than discrete concepts. 39 The reason is simple: Direct discrimination is not by definition more objectionable from the victim's viewpoint. As Moreau explains, "although there are certain heinous cases of direct discrimination . . . in most cases of direct and indirect discrimination, we can see the culpability of agents . . . as stemming from the same source: their negligence." 40 Not just that, rules amounting to prima facie indirect discrimination may be as malicious or even more malicious than cases of direct discrimination.   40 See MOREAU, supra note 16, at 184.
As the Supreme Court of Canada observed in the Meiorin case, the distinction between direct and indirect discrimination may give "employers with a discriminatory intention and the forethought to draft the rule in neutral language an undeserved cloak of legitimacy." 41 The headscarf cases bring home the point that the prohibition of direct and indirect discrimination must be seen as pursuing in essential respects the same purpose. As we saw, the employer's behaviour in Bougnaoui clearly amounted to direct discrimination, but the company policy in Achbita burdened religious people more severely. The company where Ms. Bougnaoui was employed was not opposed to employees wearing the headscarf at work and did not appear to harbour manifest prejudices towards Muslims. She had been told at the time of her appointment that wearing a headscarf might cause problems when she is in contact with customers, but she was allowed to wear one at work for almost a year and a half without problems. She was directly and unjustifiably discriminated against, but the decision to not allow her to wear a headscarf when with customers was not clearly driven by malice. It was perhaps merely due to negligence, motivated by a more innocent, though no less problematic, desire to protect customer relations. In comparison, the exclusionary effect of the "neutral" dress code in Achbita was much stronger. 42 Employees were not allowed to wear religious clothing, regardless of their function within the company and whether it involved contact with customers; and even if the dress code did not directly target the headscarf, it may have been motivated by an intention to harm Muslim women. Following Achbita, IX v. Wabe, and L.F. v. SCRL, an employer with such an intention can simply reformulate provisions that discriminate directly in neutral terms, knowing that this may license his discriminatory attitudes. 43 Thus, in light of the fact that indirect discrimination may be more burdensome than direct discrimination, it would be implausible to think that indirect discrimination necessarily involves a lesser or different kind of wrong, or that it must be subject to more lenient scrutiny just because it happens to be indirect.

II. The Implications of the Effects-Based View
The implications of the effects-based view must be spelled out carefully. First, it does not imply that the wrongness of a measure can be determined looking at its adverse effects alone; it is important not to lose sight of the justification of a measure. Consider, on the one hand, a requirement that medical staff may not wear sleeves below the elbow for hygienic reasons. Suppose it can be shown that this is an effective measure to prevent infections. It would not make sense to argue that hospitals act wrongfully by introducing such a sensible rule, merely because it adversely effects religious people who cover their forearms for religious reasons. Indeed, such rules are not morally objectionable, provided that they are an effective way of preventing infections. On the other hand, they would constitute a moral wrong if they do not contribute to the intended objective or if infections could be equally avoided by other means that do not burden religious people. Therefore, as Collins explains, "the justification element in the law of indirect discrimination . . . should be understood as an essential ingredient of the wrong." 44 It seems to me, however, that the justification element is also an essential ingredient of the wrong in direct discrimination cases. 45 Imagine the decision of a film producer to cast a white male actor for the role of Pope for reasons of authenticity. It would be odd to think that this amounts to an injustice 41 British Columbia (Public Service Employee Relations Commission) v. British Columbia Government Service Employees' Union [1999] 3 S.C.R. 3, para. 29 (Can.). 42 Even after the CJEU had decided that such a prohibition could cover only workers who interact with customers. See Achbitai, at paras. 40-43. towards non-whites or women. Here as well, determining the wrongfulness of a measure is possible only once its justification is known.
Second, it is compatible with the effects-based view to think that discrimination can be wrongful for reasons other than the adverse effects it produces. Discrimination motivated by animus or bigotry does an additional injustice to victims: Such attitudes are wrongful not only because of the adverse effects they produce but also because of the contempt and derision they put on victims. Indirect discrimination may be as reprehensible as direct discrimination, but it is not necessarily so. The effects-based view does not, therefore, need to have the radical implication that we should abandon the distinction between direct and indirect discrimination. 46 Direct discrimination may involve additional wrongs that lawmakers want to eradicate. Moreover, direct discrimination typically raises the proverbial red flag that can justify treating it differently from indirect discrimination. But as we have seen, the assumption that direct and indirect discrimination necessarily trace different wrongs, and that the wrong in indirect discrimination is less serious than in direct discrimination, rests on a mistake. Indirect discrimination can conceal the intention to discriminate and its adverse effects on victims can be just as severe. This is an important first step in understanding the central fault in the headscarf judgments. As the final section explains, the grounds for justifying indirect discrimination are broader, but the justificatory process should not be less rigorous than that of direct discrimination.

C. The Place of Religion in EU Non-Discrimination Law
To better understand the mistakes made in the headscarf case law, we need to consider not only the purpose of EU non-discrimination law, but also the reasons for including religion as a protected personal characteristic alongside other characteristics like disability and age. It is generally accepted nowadays that these characteristics must be treated the same for the purpose of non-discrimination law, 47 but, as discussed immediately below, it is occasionally assumed that mutable characteristics are less worthy of protection than immutable characteristics-an assumption that, as we shall see, seems to have influenced the headscarf judgments. Although these judgments have been criticised for this, the criticism has not been adequately substantiated. This section explains why exactly the argument set out in the previous section applies to all personal characteristics protected by EU non-discrimination law To be clear, my argument does not concern the "hierarchies" the legislative framework has created between different protected characteristics. 48 For example, because the scope of the Racial Equality Directive is broader than that of the Framework Directive, 49 EU law provides better protection against discrimination on the grounds of race and ethnicity than against discrimination based on religion or belief, disability, age, and sexual orientation. Instead of such politically motivated hierarchies, my argument concerns the belief that immutable characteristics, for example race or age, should receive better protection than characteristics that are mutable and/or a consequence of personal choice, for example religion or pregnancy. 50 This section rejects this immutability thesis and explains why the scope of the prohibition of discrimination should not depend on the ground in question. 46 For a different view, see MOREAU, supra note 16, at chapter 6. 47 For a good discussion of the shortcomings of the immutability thesis, see, Foran, supra note 23. Whether or not immutability is a relevant factor in determining the level of protection to be provided against discrimination is a question that acquired new salience after Achbita and Bougnaoui. There appears to be a mismatch between these judgments and the general body of CJEU case law, on the principle of non-discrimination, 51 which seems in part due to considerations about the personal characteristics religion and belief. The cases were subject to two contrasting opinions, by Advocates General (AGs) Sharpston and Kokott, on the place of religion in non-discrimination law. According to AG Sharpston, "it would be entirely wrong to suppose that, whereas one's sex and skin colour accompany one everywhere, somehow one's religion does not." 52 AG Kokott, on the other hand, advocated a distinction between the protected grounds, saying that the scope of non-discrimination law should depend on whether adverse treatment is based on an immutable characteristic: [T]he Court has generally adopted a broad understanding of the concept of direct discrimination, and has, it is true, always assumed such discrimination to be present where a measure was inseparably linked to the relevant reason for the difference of treatment. However, all of those cases were without exception concerned with individuals' immutable physical features or personal characteristics-such as gender, age or sexual orientationrather than with modes of conduct based on a subjective decision or conviction, such as the wearing or not of a head covering at issue here. 53 The issue of immutability was not directly addressed by the CJEU, but it is believed to have leaned towards the latter view. 54 What to make of the argument that the (im)mutability of a protected personal characteristic must influence the content and scope of non-discrimination law?
Before explaining why the immutability thesis should be rejected, let me discuss three powerful but inadequate objections that have been raised against AG Kokott's view. These objections do not provide a sufficiently strong refutation of the thesis. The first objection is that religion is not actually immutable or a chosen lifestyle. As Vickers has argued, "there is little mobility between religious groups in practice, suggesting that the exercise of any 'choice' is very limited." 55 Her point is valid, but it cannot explain why the (im)mutability of a characteristic is an irrelevant consideration. Even if people rarely alter their religion, it does not seem plausible to deny that religion and belief are mutable in a way that race, disability, or age are not. A better argument, therefore, is one that shows that it is irrelevant whether or not a protected characteristic is immutable. The second objection levelled at the immutability thesis is that it has no basis in EU non-discrimination law; it makes no mention of immutability. 56 However, since the relevant legal provisions are open ended, their meaning can be determined only by taking account of the law's purpose. The third objection that has been raised is that the application of the immutability thesis may undermine the protection of other mutable characteristics, such as sexual orientation, gender, or pregnancy. 57 This is a realistic possibility, but as with the second argument, this argument assumes what must be proven, namely that this is indeed undesirable and incompatible with the rationale of the law. Perhaps the fact that one can change one's religion or gender, or choose not to  55 See Vickers, supra note 54, at 243. 56 See Hennette-Vauchez, supra note 5, at 752-53. 57 For a similar response, see id., at 752. become pregnant, renders these grounds less protection worthy than immutable characteristics such as race or disability.
This, of course, is not the position most of us take. Normally we think that it is irrelevant, for the purposes of non-discrimination law, that pregnancy is a choice or that it is possible to change sexual characteristics through gender reassignment surgery. 58 Nor is it, contrary to what AG Kokott assumes, the position that EU law normally takes. Her claim, that a broad conception of the concept of direct discrimination has been adopted only in cases where a distinction was made on the basis of an immutable personal characteristic, is plain wrong. For instance, the CJEU has acted forcefully in disputes concerning discrimination on the grounds of pregnancy and sexual orientation. 59 Moreover, to my knowledge, no one is of the opinion that it should have taken a different stance and given less protection to pregnant women and people who were discriminated against for their sexual orientation because these are mutable characteristics. So, it seems that immutability should not be a relevant consideration-logically also not when it comes to religion and belief. Why is that so?
The reason to reject the immutability thesis is that it does not capture the wrong in discrimination. Whether or not a personal characteristic is immutable is irrelevant to whether it is protected against discrimination. Height and hair colour, for example, are not protected characteristics because they are not socially salient: they have little to no bearing on the allocation of resources and opportunities in our society. Other characteristics, in contrast, are mutable but do influence the distribution of socio-economic and political opportunities. Gender and sexual orientation clearly fall into that category, but so do religion and belief. As Laborde has observed, "religion at times functions exactly like race-an assigned identity serving to affirm and consolidate boundaries between dominant and dominated groups." 60 The protected characteristics under EU non-discrimination law are all drivers of social vulnerability, domination, and exclusion, which is why the question of whether or not they are immutable is the wrong one to ask. The immutability thesis deprives certain vulnerable social groups of the possibility to challenge the persistent disadvantages they face and must be rejected, therefore, as incompatible with the purpose of EU non-discrimination law. It should be noted right away that, as will be discussed in detail in the next section, not all personal beliefs are covered by the prohibition of discrimination, but only those that are socially salient. In that respect, the place of religion and belief within the Framework Directive should not be different from that of race or disability.

D. Direct and Indirect Discrimination on the Grounds of Religion and Belief
As has been explained, by including the characteristics of religion and belief within the scope of protection of non-discrimination law, the EU seeks to tackle the adverse effects people may experience as a result of their personal beliefs. More specifically, the Framework Directive protects religion and belief, alongside personal characteristics such as age and disability, to prevent and reduce inequalities in employment opportunities based on these characteristics. However, while it appears that the headscarf judgments misunderstood the purpose of non-discrimination law and the place of religion and belief therein, it is not yet clear where exactly their shortcomings must be located. Two lines of criticism have been raised against the judgments: (1) That neutrality policies have been wrongly classified as indirect discrimination; (2) that these policies were wrongly held to be justifiable. In this section and the next, both points of criticism are examined to clarify what 58 See also, FREDMAN, supra note 32, at 131-32.  60 See LABORDE, supra note 32, at 137. mistakes were made. In doing so, this section will seek to improve our understanding of the dividing line between direct and indirect discrimination, and the next about the conditions under which discrimination can be justified.

I. Neutrality Policies-a Hard Case
The dividing line between direct and indirect discrimination can be a serious bone of contention, although it is usually clear how a measure must be classified. 61 Direct discrimination arises when a person is treated adversely in comparison with another person in a similar situation because of a protected characteristic. 62 EU law treats the following two cases as direct discrimination: Where there is intentional discrimination on the basis of a protected characteristic, 63 and where all persons affected by a measure belong to the same protected group. 64 The disputes in Bougnaoui and Cresco Investigation are classical examples of direct discrimination on the basis of religion. In the first, an employee was discriminated against by direct reference to her Islamic headscarf; in the second, an employee was denied a day off on Good Friday because he was not a member of the church. The second case suggests that a situation of direct discrimination also occurs when the group of people advantaged by a measure belongs to the same group, Christians, even if the disadvantaged group is quite diverse. 65 In contrast, prima facie indirect discrimination arises when an apparently neutral measure puts persons with a protected personal characteristic at a particular disadvantage compared with other persons-the article will omit "prima facie" and simply say "indirect discrimination." 66 A typical case of indirect religious discrimination is a hospital rule that prohibits medical staff from wearing sleeves below the elbow-a neutral rule that has a differential impact on people who cover their forearms for religious reasons.
But where in the spectrum of discrimination does a rule prohibiting visible signs of political, philosophical, or religious belief-or, as in IX v. Wabe, "conspicuous, large-sized signs of any political, philosophical or religious beliefs" 67 -sit? In my view, this is a difficult case, 68 which falls between clear-cut cases of direct and indirect discrimination and which may explain why the reasoning in the headscarf judgments is hopelessly muddled. According to the CJEU's ruling in Achbita, such rules amount to indirect discrimination: They may cause particular inconvenience . An intention or motivation to discriminate is thus not necessary to establish direct discrimination, but it is sufficient to reach that verdict. Officially, indirect discrimination only arises when such neutral measures are not justified by a legitimate aim that is appropriate and necessary, which is why it is more correct, technically speaking, to say prima facie indirect discrimination. 67 The rule applicable to one of the applicants in that case. See, IX v. Wabe, at para. 35. 68 See also, Cloots, supra note 64. to workers following religious precepts but do not "establish a difference of treatment between workers based on a criterion that is inextricably linked to religion or belief." 69 According to IX v. Wabe, however, such rules amount to direct discrimination if they ban only conspicuous, largesized signs. Such rules are "liable to have a greater effect on people with religious, philosophical or non-denominational beliefs which require the wearing of a large-sized sign." 70 This is why the CJEU ruled that they are inextricably linked to the protected ground of religion and belief and "must be regarded as directly based on that ground." 71 However, this distinction-between rules banning all, or only conspicuous, large-sized belief signs-makes no sense. For the CJEU, a ban on conspicuous, large-sized signs, is directly discriminatory if, in practice, it has a greater effect on some religious people; yet a policy that prohibits the wearing of all visible signs does not constitute direct discrimination, even if it has a greater effect on some religious persons-and, evidently, blanket bans on religious, political, and philosophical signs have a greater effect on certain religious groups, namely people with beliefs which require the wearing of specific clothing or signs.
So the question remains: Where in the spectrum of discrimination do such rules sit? The literature unfortunately does not always offer clear answers either. Instead of recognizing that neutrality policies represent a hard case, some scholars have tried to classify them as the epitome of either direct or indirect discrimination. For Vickers, the CJEU's position that such rules in principle amount to indirect discrimination "is uncontroversial: the application of a neutral rule which has differential impact on different groups is, after all, the architype of indirect discrimination." 72 However, a rule that prohibits the wearing of visible signs of political, philosophical, and religious belief is, in a relevant sense, different from a hospital rule mandating the wearing of short sleeves-the proper archetype of indirect discrimination-if only because the former rule relies explicitly on religion and belief. Yet it is also not evident that such a rule is a clear case of direct discrimination. Here is another typical example of indirect discrimination: A rule that imposes an entirely neutral dress code that does not allow for any kind of personal view or opinion-whether religious clothing, a ribbon in support of LGBT rights, or sportswear bearing a logo of one's favourite team. 73 Is such a rule relevantly different from one that prohibits the expression of religious, political, and philosophical beliefs; and if so, why?
Neutrality policies ask us to spell out and peruse our intuitions and assumptions about direct and indirect discrimination. 74 In doing so, two objections that have been raised against the CJEU's judgment that a ban on visible signs of political, philosophical, or religious belief amounts, in principle, to indirect discrimination will be considered. These objections, called the "levellingdown objection" and the "inappropriate comparator objection," maintain that such policies amount to direct discrimination, just as a ban on Islamic headscarves alone. This discussion draws our attention to specific difficulties posed by the grounds of religion and belief and should help clarify how the boundary between direct and indirect discrimination must be drawn. The reason why these objections are unsuccessful is that they do not capture the distinction between direct and indirect discrimination. If my argument is plausible, policies prescribing religious, political, and philosophical neutrality are indeed indirectly discriminatory, unless they are motivated by an intention to discriminate. Let me consider the two objections in turn. 69 Achbita, at paras. 30-32. See also, IX v. Wabe, at para. 53. This example is borrowed from, Achbita, Opinion, at para. 110. 74 See also, Sharpston, supra note 9, at paras. 201-06.

II. The Levelling-Down Objection
The levelling-down objection is, as the name implies, opposed to resolving discrimination by levelling down, that is, by taking an advantage away from the advantaged group rather than giving it to the discriminated group. Thus, discrimination against Muslims should be solved not by removing an advantage from other (non-)religious people, but by extending that advantage to Muslims. However, critics of the headscarf cases have argued not just that levelling down is wrong, but also that it amounts to direct discrimination. More specifically, they say that direct discrimination does not cease to be direct when discrimination is addressed by levelling down. As the referring court in IX v. Wabe put it: "the difference between indirect and direct discrimination can be conditioned not by whether other people with other (protected or unprotected) characteristics are also treated similarly poorly." 75 It indeed seems problematic to prefer a rule that harms all religious groups to one that affects Muslims alone. This intuition finds support in our conclusion in Section B, that apparently neutral rules can be more harmful to religious groups than directly discriminatory ones. Moreover, as Khaitan makes clear: "a levelling-down strategy adopted mainly to avoid extending the benefit in question to a protected group is likely to have an expressive adverse effect on this group." 76 Companies that adopt neutral dress codes because they seek to deny Muslims the right to wear religious clothing reinforce their inferior status: Their policies have an expressive adverse effect on Muslim employees. Now recall what the CJEU decided in Bougnaoui: Employers are not permitted to institute a ban on the Islamic headscarf alone, but they can avoid breaching the prohibition of discrimination by adopting a blanket ban on all visible signs of political, philosophical, and religious belief. 77 This conclusion is egregious because it invites company policies causing expressive harm to religious minorities.
However, the levelling-down objection rests on an inadequate conceptual understanding of the distinction between direct and indirect discrimination. Nor does it offer reasons to believe that a company rule prohibiting the manifestation of a wide range of personal beliefs is directly rather than indirectly discriminatory. As we will see in the next section, it does explain why the justificatory burden for such a rule is hard to meet. Consider the Barber ruling, in which the CJEU found that a lower pension age for women than for men amounted to direct sex discrimination. 78 Quite predictably, the ruling hardly benefited men: Instead of lowering their pension age, UK pension funds "levelled down" by raising women's pension age to that of men, in a policy change later sanctioned by the CJEU. 79 However, despite treating women equally poorly as men, and not in any way redressing the structural disadvantages faced by women, 80 this policy was not directly discriminatory against women. In other words, levelling down as such is not sufficient to establish direct discrimination. Now let's consider what is probably a more controversial case. A company that employs some people who wear religious clothing at work wishes to project an image of complete neutrality. Knowing that it cannot ask its employees to remove their religious dress, the company introduces a uniform to be worn by all employees. However, while such a uniform clearly makes all employees worse off, it is generally accepted that such dress codes are not directly discriminatory. 81 Of course, they will if the company's decision to level down is predicated on stereotypes and prejudices, and they may be indirectly discriminatory if no reasonable 75 IX v. Wabe, (emphasis added). The same objection has been raised by, Hennette-Vauchez, supra note 5, at 748. accommodations are made for religious people. 82 Still, as this example suggests, levelling down alone is insufficient to support the finding of direct discrimination.

III. The Inappropriate Comparator Objection
One might object that asking all employees to wear a uniform is relevantly different from a rule prohibiting manifestations of political, philosophical, and religious belief. The former affects all employees whereas the latter affects only those with certain personal beliefs. Unlike the first rule, that is, the second rule seems to differentiate directly based on the characteristics of religion and belief. 83 This objection has merit but should be distinguished from the levelling-down objection. The objection, rather, is that a rule prohibiting the manifestation of political, philosophical, and religious beliefs is a clear instance of direct discrimination, since it adversely affects employees who have, or manifest, a particular religion or belief, when compared to employees who do not have, or manifest, a particular religion or belief.
For the CJEU, the relevant comparator was employees who want to manifest their beliefs. It said that a neutrality policy may only amount to indirect discrimination because it "covers any manifestation of such beliefs without distinction." 84 According to some scholars, this is the wrong conclusion, using the wrong comparator. Cloots has argued that "the groups which should have been compared . . . are not those of religious and non-religious people but, rather, those of people who do and do not manifest their religious faith in public." 85 Similarly, according to Hennette-Vauchez, "a situation that entails discrimination on the grounds of religious beliefs should not only be assessed by comparing it to the treatment of persons expressing other beliefs, but also to persons expressing no belief." 86 Thus, according to the inappropriate comparator objection, religious discrimination is direct not just when it is directed against one or a select group of beliefs -as in Bougnaoui and Cresco Investigation-it is also direct when people who manifest their beliefs are treated unfavorably compared to those who do not, or more generally, as Cloots explains, when "all people who incur disadvantage are members of the protected group," 87 which would be so in this case because the rule affects only people with a belief and not those without. Brems underlined the latter point by drawing a comparison with the protected ground "disability": "it seems unthinkable that the ECJ would rule that a measure excluding without distinction persons with all kinds of disabilities would not constitute direct discrimination." 88 Before examining the inappropriate comparator objection, it is worth clarifying what seems to be the most plausible interpretation of the CJEU rulings in relation to the scope of the protected grounds of religion and belief. For Cloots, the decision that the concept of religion encompasses both "the forum internum, that is the fact of having a belief, and the forum externum, that is the manifestation of religious faith in public," 89 entails that the public manifestation of one's belief has become a protected characteristic on its own terms. 90 Therefore, she argues, the group of people disadvantaged by neutrality policies coincides entirely with the protected group of people who 82 This can also be read into Achbita, supra note 1, at paras. 42-43, although the CJEU set the bar too low. The justification for indirect discrimination is discussed in the final section. 83 For such an argument, see Eva Brems, European Court of Justice Allows Bans on Religious Dress in the Workplace IACL-AIDC BLOG (Mar. 26, 2017), https://blog-iacl-aidc.org/test-3/2018/5/26/analysis-european-court-of-justice-allows-bans-onreligious-dress-in-the-workplace. 84 Achbita, supra note 1, at para. 30; IX v. Wabe, supra note 2, at para. 52; L.F. v. SCRL, supra note 2, at para. 33. 85 Cloots, supra note 64, at 601. 86 Hennette-Vauchez, supra note 5, at 747. Similar points of criticism have been raised by, Weiler, supra note 5, at 1005-006; Vickers, supra note 54, at 250. 87 Cloots, supra note 64, at 605. 88 Brems, supra note 83. 89 Achbita, supra note 1, at paras. 28; IX v. Wabe, supra, note 2, at para. 45. 90 See Cloots, supra note 64, at 601. manifest their religion, meaning that the rule is a typical case of direct discrimination. 91 However, another less-controversial interpretation of the CJEU's reasoning is possible, namely that the manifestation of a belief is covered by, and indeed related to, the characteristic of religion and belief just as having a belief is. That the manifestation of religious beliefs is covered by the Framework Directive is undisputed, and the judgments merely seem to implement this undisputed fact. Hence, if the CJEU erred by not finding direct discrimination, it is not because it decided that the manifestation of one's belief is a distinct protected characteristic. Now let us turn to the inappropriate comparator objection. It is a powerful objection, since it seems inconceivable that a measure that relies directly on a protected personal characteristic does not amount to direct discrimination, even more so since neutrality policies seem to adversely affect only people who are members of a protected group. In this respect, such policies seem comparable to employment policies that exclusively disadvantage pregnant people, which amount to direct sex discrimination since only women can be affected by such policies. 92 And yet, as will be explained now, there is a reason to think that this may not be the case. To understand this, we must explore some of the specific challenges posed by the grounds of religion and belief.
There is an important difference between, on the one hand, the protected grounds of religion and disability and, on the other hand, the protected grounds of sex, age, and race. The latter are characteristics that all individuals have-no one is without a particular sex, age, or race. For this reason, as Khaitan observed, "the fundamental unit of protection qua grounds exists in its particular order (femaleness) rather than in the universal order (sex)." 93 It is possible to discriminate against women, or men, but not to discriminate against sex (i.e. all individuals). But religion and disability, as well as pregnancy, are different. Not everyone is religious, just as not everyone has a disability or the possibility of becoming pregnant. 94 And it is possible to discriminate against pregnant women and disabled people as compared to people who are not pregnant or have no disability, just as it is possible to treat religious people less favorably in comparison with non-religious people. Imagine a militantly atheist employer dismissing its employees if they are found to adhere to a particular religion. Clearly this would be an instance of direct discrimination on the basis of religion, despite the fact that the employer does not distinguish between different beliefs.
At the same time, religion is not alike disability and pregnancy in every respect. Religion, as we just saw, takes a symmetric approach: It also protects people who have no religion. For instance, non-religious persons discriminated against by their religious employer can claim religious discrimination. 95 Disability, in contrast, enjoys asymmetric protection: A defining factor of discrimination law across the world is that "only a disabled person can bring a claim of discrimination; an able-bodied person cannot claim disability discrimination." 96 Pregnancy is more complex, as it falls under the protected ground of sex in EU non-discrimination law, 97 but pregnancy as such is also asymmetric: Only women can claim pregnancy discrimination. 98  Khaitan, supra note 12, at 30-31. See also Brems, supra note 83. 95 Egenberger, supra note 6; IR v. JQ, supra note 6. For discussion, see van den Brink, supra note 6. 96 Khaitan, supra note 12, at 61. Dekker, supra note 92, at para. 12. 99 IX v. Wabe, supra note 2, at para. 52; see also, L.F. v. SCRL, supra note 2, at para. 34. is possible to discriminate against the universal order religion as well as particular instances of this universal order, such as Islam, Christianity, Jewism, and so forth.
But what really complicates the question of whether a neutrality policy is directly or indirectly discriminatory is the breadth of the protected ground "belief." We are considering a rule precluding not only the wearing of religious signs such as the Muslim headscarf or the Jewish kippah, but also of clothing and signs manifesting someone's political affiliation or philosophical convictions. It seems counterintuitive to treat all these views as equivalent for the purpose of nondiscrimination law. Asking employees not to wear their Muslim headscarf is, instinctively at least, very different from asking them not to wear clothing showing their political leanings or their commitment to a particular lifestyle. Religious beliefs are protected because they are socially salient: They affect the distribution of social and economic advantages because religious people are susceptible to stigmatization, stereotyping, and humiliation. 100 The inclusion of religion within the scope of the prohibition of discrimination is intended to remedy these disadvantages and protect the integrity of religious individuals, i.e. their ability to live in accordance with their deepest commitments-commitments which "cannot be sacrificed without feelings of remorse, shame, or guilt, by contrast to preferences, which can." 101 In contrast, other personal beliefs may be trivial and peripheral to our sense of identity, akin to preferences rather than deep commitments, and capable of being sacrificed at no significant personal cost.
Of course, people are also stigmatised and prejudiced for non-religious beliefs, which is why such beliefs, including the absence of any belief in deities, can be socially salient as well. Of course, social salience is a matter of degree: Religious beliefs may structure social interactions to a greater extent than most non-religious beliefs. But this does not mean that the latter do not substantially affect the distribution of socio-economic advantages; that they are not socially salient. This explains why belief is a protected characteristic alongside religion; not just under EU nondiscrimination law, 102 but also, for instance, under Article 9 of the European Convention of Human Rights.
But not every personal belief is protected for the purpose of non-discrimination law. Both legal scholarship and legal doctrine typically assume that non-religious beliefs must be analogous to religious beliefs to enjoy protection. 103 According to the case law of the European Court of Human Rights, for example, only "views that attain a certain level of cogency, seriousness, cohesion and importance" are covered by the Convention. 104 Likewise, German courts have found that the concept of belief must be oriented towards religion, 105 just like UK courts, who understand the concept belief to refer to "philosophical beliefs" such as environmentalism and spiritualism. 106 In IX v. Wabe, the CJEU took the same view, ruling that "the terms 'religion' and 'belief' must be analysed as two facets of the same single ground of discrimination . . . [T]he ground of discrimination based on religion or belief . . . covers both religious beliefs and philosophical or spiritual beliefs. 107 Beliefs such as atheism, pacifism, and humanism will thus be protected by EU non-discrimination law. BVerwG, Apr. 18, 2018, 28 K 6.14, para. 58; BVerwG, July 7, 2004, 6 C 17.03. 106 See Vickers, supra note 103, at 24-25; see also Explanatory Notes to the UK Equality Act 2010, Section 10 (UK). 107 IX v. Wabe, supra note 2, at para. 47; L.F. v. SCRL, supra note 2, at para. 27.
Equally important, however, is what sorts of personal beliefs are not covered by the prohibition of discrimination; political beliefs. According to the CJEU, "the ground of discrimination based on religion or belief is to be distinguished from the ground based on 'political or any other opinion.'" 108 National legislation and the case law of domestic courts on the status of political convictions under non-discrimination law are not consistent, 109 but the position taken by the CJEU seems compatible with that of other EU institutions: Political opinions do not fall within the concept belief under EU non-discrimination law. National authorities may, of course, adopt a broader understanding covering political beliefs. 110 The exclusion of political beliefs from the scope of the law does not avoid the difficulty of definition-indeed, as Vickers notes, "it just shifts location. The divide is located between 'religious or philosophical beliefs" which are subject to special protection, and other beliefs which are not.'" 111 This does not mean that courts face serious difficulties in determining whether someone's personal beliefs are covered by the prohibition of discrimination: Disputes about the proper definition of religion and belief arise only at their margins. 112 For example, whether socialism or Marxism are prohibited grounds for discrimination depends on whether they are to be understood as a political or philosophical belief, 113 and thus on whether they attain the appropriate level of cogency, seriousness, cohesion, and importance. However, personal other beliefs that are incomparable to religious beliefs, such as support for a political party, are in principle not protected under EU non-discrimination law.
This divide between religious and philosophical beliefs, on the one hand, and political and other beliefs, on the other, not just has the advantage of being broadly compatible with the purpose of EU non-discrimination law. It is also broadly aligned with the way in which the two protected grounds are interpreted at national level and, importantly, supports our intuition that there is a relevant difference between asking employees not to wear their Muslim headscarf and asking them not to wear clothing showing their political affiliation or commitment to a particular lifestyle. Banning the manifestation of personal beliefs is far less objectionable if they can be given up without incurring non-trivial costs such as feelings of embarrassment or guilt. Here are some examples. I might have a weak claim to wear the apparel of my preferred rock band, while my religious colleagues have a strong right to wear a headscarf, turban, or kippah. My politically conscious colleague might have a weak claim to wear a tie of his preferred political party, while my vegan colleague has a strong claim not wear company clothing made with animal products. And my football-obsessed friend might have a weak right to watch her local football team play during working hours, while her religious colleague has a strong right to observe her praying rituals while at work. 114 Why is all this relevant to the definition of direct and indirect discrimination and, more specifically, to our understanding of the inappropriate comparator objection? According to this objection, a policy treating employees who have, or manifest, a religion or belief less favorably than those who do not have, or manifest, a religion or belief amounts to direct discrimination even where there is no discernible intention to cause harm, simply because such measures exclusively 108 IX v. Wabe, supra note 2, at para. 47; L.F. v. SCRL, supra note 2, at para. 27. affect members of the protected group religion or belief. Employers can intentionally invoke the protected characteristics of religion or belief, acting out of malice or bigotry, in which case their behavior constitutes direct discrimination. But employers can also institute rules covering all personal beliefs without any intention of harming persons with a belief. Consider a rule prohibiting judicial officials from manifesting their religious, political, and philosophical beliefs. It seems evident that such a neutrality policy is not motivated by an intention to harm people based on their religion or belief. The question is whether, as the inappropriate comparator objection suggests, such a policy nevertheless amounts to direct discrimination.
We can now understand that this is most likely not the case. Direct discrimination occurs when a group of persons affected by a rule consists entirely of members of a protected group. 115 Neutrality policies do not have this effect: They adversely affect personal beliefs that are protected from discrimination, including religious beliefs, and those that are not, including political beliefs. And because the group of persons adversely affected by the policy does not coincide entirely with members of a protected group, it does not fall within the prohibition of direct discrimination. 116 This shows that a ban on the manifestation of political, philosophical, or religious belief is very similar to a rule that prohibits the manifestation of any form of personal belief, including a person's allegiance to a particular sports club, university, or musical tastes. Such a rule is also closely related to the protected ground of 'belief' but is a classic example of potential indirect discrimination. This will be different only if a neutrality policy is adopted out of malice or prejudice, with the intention to discriminate. Only then does such a rule amount to direct discrimination.
To sum up, this section considered two objections to the CJEU's decision that neutrality policies amount to indirect and not direct discrimination. Both objections were found not to properly capture the distinction between direct and indirect discrimination. In principle, such neutrality policies seem to amount to indirect rather than direct discrimination, except for policies motivated by an intention to discriminate. As the foregoing shows, however, the distinction is not always easy to establish, which explains why the classification of a rule as direct or indirect discrimination cannot be conclusive in determining its lawfulness. Imposing an indirect burden on religious adherents is not acceptable unless justified on legitimate grounds, whereas a provision that directly burdens religious beliefs is acceptable in certain situations. Ultimately, therefore, the crux of the legal analysis in discrimination cases is whether a discriminatory provision can be justified.

II. Justifying Religious Discrimination
Whether a rule amounts to direct or potential indirect discrimination is relevant to assess which justifications can be accepted. The available justifications for a directly discriminatory rule are limited and specifically listed in the Framework Directive. The main justification is the exception for occupational requirements, laid down in Article 4(1) of the Framework Directive. 117 For indirectly discriminatory rules, the available justification is broader. According to Article 2(2)(b) 115 Note, in this respect, that one can be a member of a religious without adhering to that religion. See Tarunabh  Article 4(2) provides for an occupational requirement exception specifically for religious employers, a subject which is beyond this article's scope. Two other justifications for direct discrimination will also not be discussed: As we saw in section 2, Article 7 authorises the adoption of positive measures to prevent or compensate for disadvantages linked to a protected characteristic. Finally, there is Article 2(5) which provides that the Directive is 'without prejudice to measures laid down by national law which, in a democratic society, are necessary for public security, for the maintenance of public order and the prevention of criminal offences, for the protection of health and for the protection of the rights and freedoms of others.' of the Framework Directive, an apparently neutral rule which puts persons of a particular religion or belief at a particular disadvantage does not amount to indirect discrimination when it "is objectively justified by a legitimate aim and the means of achieving that aim are appropriate and necessary." This final section will examine both justifications. The application of Article 2(2)(b) of the Framework Directive in Achbita and IX v. Wabe has been strongly criticized, but the validity of such criticism can, in my view, be properly understood only by comparing its application with that of Article 4(1) of the Framework Directive. A comparison of the CJEU's attitude towards both provisions will show that the bar for justifying discrimination under Article 4(1) of the Framework Directive was set much higher than under Article 2(2)(b) of the Framework Directive: Catering to the prejudiced views of customers was not legitimate within the scope of the former, while it is within the scope of the latter. This contradiction seems motivated by the view that direct discrimination constitutes a more serious form of discrimination than indirect discrimination, which would explain why the CJEU was too permissive in examining whether the conditions in Article 2(2)(b) of the Framework Directive had been met. The main problem in the headscarf cases is essentially this: What is not a legitimate aim within the meaning of Article 4(1) of the Framework Directive can be within the meaning of Article 2(2)(b) of the Framework Directive.

III. Justifying Direct Discrimination
The occupational requirement exception in Article 4(1) of the Framework Directive provides that, Member States may provide that a difference of treatment which is based on a characteristic related to [a protected ground, including religion or belief] shall not constitute discrimination where, by reason of the nature of the particular occupational activities concerned or of the context in which they are carried out, such a characteristic constitutes a genuine and determining occupational requirement, provided that the objective is legitimate and the requirement is proportionate.
The Directive clarifies that the exception applies in "very limited circumstances." 118 First of all, Member States must have provided for this exception in their national legislation. Where they have done so, employers may differentiate on the basis of a characteristic related to a protected ground on the condition that it constitutes a "genuine and determining" occupational requirement and that it pursues a "legitimate objective that is proportionate." The CJEU has understood this provision in a manner consistent with legislative intent; it has decided that it must be "interpreted strictly." 119 The Bougnaoui case raised questions about the correct interpretation of the derogation. Ms. Bougnaoui was dismissed for refusing to remove her Islamic headscarf when she was with customers of her company. The question before the CJEU was whether the customer's wish not to have services provided by someone wearing the Islamic headscarf could constitute a genuine and determining occupational requirement. AG Kokott argued in her Opinion in Achbita that it could, and that a company, in pursuit of its business interests, can use Article 4(1) of the Framework Directive to require its employees "to behave and dress in a particular way at work company policy," especially employees who are in "regular face-to-face contact with customers." 120 The CJEU disagreed and decided that an occupational requirement for the purpose of Article 4(1) of the Framework Directive must be "objectively dictated by the nature of the occupational activities concerned [and cannot] cover subjective considerations, such as the willingness of the employer to take account of the particular wishes of the customer." 121 This view is correct, but the CJEU did not adequately explain why Article 4(1) of the Framework Directive cannot be used to satisfy the subjective wishes of customers. To do so, we must analyse the structure of the provision in greater detail.
All employers may, in principle, avail themselves of this provision. First, the exception can be used by employers with an ethos based on religion or belief to justify discrimination against nonadherents in respect of employment activities that support their religious mission. For example, the Catholic Church can discriminate against non-Catholics in the appointment of clergy. 122 When religious employers can rely on this provision is somewhat context-dependent, but Vickers' interpretation of Article 4(1) of the Framework Directive seems accurate: "religious discrimination is only really likely to be lawful in the case of those employed in religious service, whose job involves teaching or promoting the religion, or being involved in religious observance." 123 Second, non-religious employers can also invoke Article 4(1) of the Framework Directive to use religion or belief as an occupational requirement. Imagine an advertiser's decision to reject a candidate for modelling in a shampoo ad because she insists on wearing the Islamic headscarf. Religion is not a defining feature of the modelling job in question, but it is absolutely necessary that the model is willing to uncover her hair. Thus, in this instance, the decision not to hire a model who refuses to remove her headscarf would amount to a difference in treatment based on a characteristic related to religion that constitutes a genuine and determining occupational requirement.
Although non-religious employers may rely on Article 4(1) of the Framework Directive to justify religious discrimination, the employer in Bougnaoui had not satisfied the conditions for the application of this provision. It is generally accepted that it sets the bar for justifying differences of treatment on the basis of religion or belief very high. 124 As AG Sharpston rightly remarked, the words "genuine" and "determining" suggest that "the derogation must be limited to matters which are absolutely necessary in order to undertake the professional activity in question." 125 A decision to reject someone who refuses to unveil her hair to act in a shampoo commercial would clearly meet these conditions. So would a hospital rule imposing a dress requirement on personnel for health and safety reasons (this rule would also meet the justification for indirect discrimination). In contrast, neutral clothing rules do not seem strictly necessary for the exercise of most professions. Ms. Bougnaoui worked as a design engineer and could meet her job requirements with or without a headscarf. It thus appears that the CJEU was correct to find that "the concept of a 'genuine and determining occupational requirement'.. . refers to a requirement that is objectively dictated by the nature of the occupational activities concerned [and cannot] cover subjective considerations." 126 However, the CJEU did not explain correctly why the conditions in Article 4(1) of the Framework Directive had not been met. One could argue that being able to meet customers' 120 See Opinion of AG Kokott, supra note 53, at paras. 79-83. 121 Bougnaoui, supra note 1, at para 40. Following the Opinion of AG Sharpston, supra note 52. 122 Article 4(2) provides a broader exception specifically for employers with an ethos based on religion or belief, so Article 4(1) will be particularly relevant to such employers if they operate in Member States that have not incorporated the exception laid down in Article 4(2) into their domestic law. This includes Sweden and France: Country Report on Non-Discrimination: Sweden (95) 40; European Equality Law Network, Country Report on Non-Discrimination: France, (2020) 69, https://www. equalitylaw.eu/downloads/5260-france-country-report-non-discrimination-2020-pdf-1-77-mb. 123 See Vickers, supra note 103, at 145; see also, McCrea, supra note 100, at 162; see also van den Brink, supra note 6. 124 Even by Opinion of AG Kokott, supra note 53, at para. 78, who however wants to lower this bar significantly. 125 Opinion of AG Sharpston, supra, note 52, at para. 96. 126 Bougnaoui, supra note 1, at para. 40. subjective wishes is an occupational requirement that is genuine and determining. Consider the example Lippert-Rasmussen provides: Suppose that a shopkeeper defends his decision not to hire any minority members on the ground that while he personally would not mind employing them, he believes-correctly, let us suppose-that many of his customers prefer being served by a nonminority shop assistant. He is in business to earn a profit, so from his point of view majority membership is a relevant ground for differential treatment. 127 Suppose further that this ground is relevant in a way that satisfies the "genuine" and "determining" conditions in Article 4(1) of the Framework Directive; that being a nonminority employee is absolutely necessary in order to fulfil the occupational activity in question. 128 Logically speaking, then, the decision not to hire a minority member is in such situations based on a requirement that is genuine and determining.
For this reason, it matters that Article 4(1) of the Framework Directive lists two additional conditions that must be met: An occupational requirement must be genuine and determining, but it must also pursue an objective that is "legitimate" and "proportionate." 129 Legitimate objectives include public security, 130 public safety, 131 and the operational capacity of public services such as police or fire services. 132 Occupational requirements are proportionate if they are sufficiently tailored. 133 However, the desire to accommodate the prejudiced wishes of customers cannot be considered a legitimate objective, let alone a proportionate one. 134 We would never allow a company to invoke Article 4(1) of the Framework Directive to justify a decision not to employ people because of their race or ethnicity in order to satisfy the bigoted wishes of its customers. 135 Even if such an occupational requirement would be genuine and determining, it would not pursue a legitimate objective for the reason that the objective pursued is contrary to the very purpose of EU non-discrimination law. The occupational requirement exception cannot be used to justify occupational requirements that pursue such bigoted objectives. For the same reason, it should not be relied upon to justify discrimination against religious people for the purpose of maximizing profit.

IV. Justifying Indirect Discrimination
This section focuses on the general justification for indirect discrimination in Article 2(2)(b) of the Framework Directive. 136 A measure will not amount to indirect discrimination if it "is objectively justified by a legitimate aim and the means of achieving that aim are appropriate and necessary." In the headscarf cases, the CJEU ruled that a neutrality policy may be regarded as legitimate, especially if the "employer's wish to project an image of neutrality towards customers relates to the freedom to conduct a business that is recognised in Article 16 of the Charter." 137 Such a policy is objectively justified, however, "only where there is a genuine need on the part of that employer, 127 Lippert-Rasmussen, supra note 18, at para. 24. See also, Hennette-Vauchez, supra note 5, at 754-57. 135 See, by analogy, Feryn.

136
Framework Directive, art. 2(2)(b). 137 IX v. Wabe and MH Müller Handels, supra note 2, at para. 63; L.F. v SCRL, supra note 2, at para. 39. See also Achbita, supra note 1, at para. 38. which it is for that employer to demonstrate." 138 In this regard, account may be had for "the rights and legitimate wishes of customers or users." 139 Furthermore, the policy must be "strictly necessary in view of the adverse consequences that the employer is seeking to avoid," 140 which requires, in particular, that the policy applies only to "those workers who are required to come into contact with the employer's customers." 141 A major criticism of the headscarf cases has been that that Article 2(2)(b) Framework Directive was not properly enforced. Especially after Achbita, it was said that the CJEU had been too lax in its assessment of the justification of neutrality policies. 142 Although it has sought to clarify and partly improve its justification analysis in subsequent judgments, including by saying that its "interpretation is inspired by the concern to encourage, as a matter of principle, tolerance and respect, as well as acceptance of a greater degree of diversity," 143 these judgments have also been met with criticism. 144 Rightly so, as will be explained, but not for the reasons many have given.
Critics have vented their disbelief at the CJEU's half-hearted application of the principle of proportionality, in particular at the fact that it stopped short of applying the principle of strict proportionality. 145 As is well known, a proportionality analysis consists of three subtests: First, the means chosen must be appropriate to furthering the aim pursued (appropriateness). Second, the means chosen must be necessary to furthering that aim (necessity). Third, the benefits of the measure must be proportionate to the damage caused to the rights and liberties of individuals (strict proportionality). 146 Neither in Achbita nor in IX v. Wabe and L.F. v. SCRL did the CJEU apply this third test, assessing whether the harm caused to Muslim employees by a neutral dress code was proportionate to the benefit obtained by a company in protecting its image of neutrality.
The appeal of this criticism is obvious. It seems inconceivable that the CJEU would pay no attention at all to whether the interests of the company outweigh those of individual employees. However, the criticism overlooks a rather basic fact, namely that Article 2(2)(b) of the Framework Directive does not provide for an application of strict proportionality. This provision merely provides that measures pursue a legitimate aim and that they must be appropriate and necessary. Thus, what seems inconceivable can be explained on the basis of a close reading of the text of Article 2(2)(b) of the Framework Directive: Strict proportionality is not among the requirements listed there, so it was not applied.
This may not render the criticism moot, but advocates of the application of the principle of strict proportionality must do more to justify their position. They have put forward two arguments, neither of which is fully convincing. The first is that an application of this principle is consistent with CJEU case law. 147 However, most discrimination disputes are resolved without. As Kilpatrick has observed in her analysis of indirect discrimination on grounds of sex case law, "the Court has not carried out strict proportionality analysis on measures found to be necessary or 138 objectively justified." 148 The second argument is that the principle of proportionality is a general principle of EU law, also enshrined in Article 52 of the Charter of Fundamental rights, and must therefore be applied when interpreting Article 2(2)(b) of the Framework Directive. However, the proposition that legislation must be interpreted in accordance with principles of EU primary law cannot not easily be justified. The EU legislature already considered and struck a balance between various competing rights, interests, and principles when it enacted the Framework Directive. It would be improper for the CJEU to perform this task again, possibly arriving at a different balance than the legislature intended. 149 Of course, discrimination disputes generally involve a balancing of competing rights-for example the right to be free from discrimination versus the right to religious freedom, 150 or interests-for example the right to be free from discrimination versus an employer's financial interests, but this ideally takes place within the confines of the legislative framework, primarily at the justification stage. In this context, it should be recalled that the CJEU interpreted the Framework Directive in accordance with the right to freedom to conduct a business in Article 16 of the Charter in Achbita. This approach has rightly been criticized for weakening the protection afforded by the Framework Directive. 151 It seems inconsistent to reproach the CJEU for applying one constitutional principle and not another.
Those who feel uncomfortable with the CJEU's conclusions should instead focus their criticism on the first element of the justification analysis-the legitimate aim assessment. This hurdle is generally easy to clear: It is legitimate to protect and pursue interests such as health and safety, 152 public security, public order, and the rights of others, 153 and sometimes business interests as well. 154 Yet the neutrality policies in question should have stranded at this hurdle: They do not pursue a legitimate aim. Indeed, the judgments' main deficiency is that the concept of a legitimate aim was construed so broadly as to include, as the CJEU held, "the desire to display, in relations with both public and private sector customers, a policy of political, philosophical or religious neutrality." 155 To see this, we must understand that not all neutrality policies should be treated the same way: Neutrality is a secondary aim to that of several possible alternative aims. For instance, the judiciary may be expected to adhere to an image of neutrality to avoid the appearance of partiality, and some Member States do not allow public school teachers to wear religious clothing to provide pupils with a learning environment free from religious influence. Whether such measures are legitimate and necessary remains disputed, but the question here is what aim private employers pursue in promoting an image of neutrality in relations with their customers. Indeed, as AG Sharpston aptly put it, it is unclear what private employers need a policy of neutrality for, were it not their interest to increase profit by "pandering to prejudice." 156 Seeing that it ruled that the wish to project an image of neutrality is legitimate where it involves "only those workers who are required to come into contact with the employer's customers," 157 the CJEU understood this very 148 well. Yet it failed to grasp the problem, tolerating policies that are adopted for no other reason than to avoid offending prejudiced customers.
In my view, this can only be explained by the apparent view that direct discrimination is a more serious offence than indirect discrimination and should, therefore, be subject to stricter scrutiny. Above we saw that catering to the prejudiced views of customers is not a legitimate aim within the meaning of Article 4(1) of the Framework Directive. The CJEU did not allow employers to discriminate directly against people wearing the Islamic headscarf on the ground that their customers take offense thereto. Similarly, it ruled in the past that employers may not discriminate against black or LGBT people because their customers take offense to their skin colour or sexual orientation. 158 Is it any less revolting to pander to the prejudices of customers based on measures that are indirectly discriminatory? To ask the question is to answer it. Policy objectives that are antithetical to the basic rationale of non-discrimination law should not be considered legitimate, irrespective of whether they are pursued by measures that amount to direct or indirect discrimination. 159 Any other conclusion misses the very point of the law.
Let me consider two possible objections to this conclusion. First, someone may argue that the European Court of Human Rights (ECtHR) has come to a very similar conclusion as regards the right of private employers to ban religious clothing. In Eweida, it said that an employer's wish to project a certain corporate image is "undoubtedly legitimate." 160 It is worth bearing in mind, however, that it said so in response to a uniform worn by employees. As Sharpston has observed, "where an employer does want to ensure that its employees project its corporate image, the standard way for it to do so is to create a uniform for them to wear." 161 Moreover, also here we must go on to ask what the aim is of pursuing a certain corporate image. In principle, an employer may prescribe a uniform so that its employees are recognizablyor representatively dressed. But should we also accept a corporate image as legitimate if it is intended to prejudice religious minorities or any other group protected by law from discrimination? One hopes that the ECtHR would not accept this; the CJEU certainly should not for the reasons just given.
A second possible objection to my conclusion would draw our attention to Article 16 of the Charter of Fundamental Rights, which guarantees employers the right to freedom to conduct their business. Someone may argue that this provision allows companies to choose their corporate ethos and image. However, besides the fact that, as explained above, this reasoning seems inconsistent with the legislature's intention in the Framework Directive, this objection poses the following question: Should the CJEU also allow employers to invoke their freedom to conduct a business to pursue a corporate ethos that is racist or homophobic? The case law on racial or sexual orientation discrimination never invokes Article 16 of the Charter to free employers from the prohibition of discrimination, and rightly so. The Framework Directive represents the legislative intention to restrict the freedom to conduct a business by a robust framework for combating discrimination. This framework should not be watered down using Article 16 of the Charter, neither in the case of racial and sexual orientation discrimination, nor in the case of religion and belief discriminationrecall section 2: The same protection should be afforded to different protection grounds). 162 The right to freedom to conduct a business should not be applied to license prejudicial practices.
The fact that the company rule in Achbita survived the first stage of the justification analysis explains why the subsequent stage of the analysis, the assessment of the rule's necessity was, quite 158 The latter follows clearly from the judgments in Feryn, supra note 134; Asociaţia Accept, supra note 59. 159 For the same conclusion, see, Brems, supra note 84; Jolly, supra note 146, at 311-12; Loenen, supra note 73, at 65; Sandhu, supra note 5; Cloots, supra note 64, at 613. 160 Eweida, et.al, supra note 104, at para. 94. The CJEU invoked Eweida to justify its own decision, see Achbita, supra note 1, at para. 39. 161 Shadow Opinion of former AG Sharpston, supra note 9, para. 130. 162 Contra, Sharpston, supra note 9, at paras. 63-80 (who takes the view the CJEU should have balanced Article 16 of the Charter against other Charter rights. This balancing approach is intellectually flawed as it still threatens to undermine the intention of the Framework Directive). frankly, embarrassing. The CJEU held that a ban on wearing visible signs of philosophical, political, or religious belief meets the condition of necessity provided that it "covers only G4S workers who interact with customers." 163 It ruled that employers must first seek to offer a religious employee "a post not involving any visual contact with those customers, instead of dismissing her." 164 As Weiler has pointed out, the CJEU essentially told religious minorities, you are fine "provided you keep out of sight, conceal your identity and your religion and do not come into contact with us." 165 Surely we would not consider it appropriate to demote black or transgender persons to a position without visual contact with customers in order to satisfy the preferences of the latter. This is just as unacceptable where religious people are concerned. The problem, however, is not the application of the necessity test, but rather the prior conclusion that private employers can legitimately pursue a policy of neutrality in order to satisfy customers. If a policy intended to satisfy the prejudiced views of customers is legitimate, it will be strictly necessary if its application is limited to employees who come into contact with customers. The only way to avoid such unpalatable conclusions is to question whether private employers can pursue an image of neutrality at all.
Finally, a company may argue that it pursues a neutrality policy not to satisfy the wishes of its customers but with the aim to avoid social conflicts at work. 166 Avoiding social conflict in the workspace is no doubt a legitimate aim, but not one that can justify banning all religious clothing at work. It would be legitimate to prohibit employees from proselytizing for their religious or philosophical beliefs, which may indeed be a source of conflict. However, a Muslim woman wearing a headscarf observes her religious beliefs; she is not proselytizing for them. This difference is crucial. 167 A company may have good reasons to not allow employees to impose their religious views on colleagues or to convert them to their faith, in order to avoid social conflict, but a rule that bans all expressions of religious, philosophical, and political belief is not a necessary means to realize that aim. If the aim is to avoid social conflict, the rule should be no proselytizing rather than no religious clothing. In a final attempt to justify its neutrality policy, a company may argue that it helps to avoid social conflict because certain employees take offense to religious clothing such as the headscarf or kippah. The response to that should be clear: It is impermissible to accommodate the wishes of prejudiced colleagues, just as it is wrong to cater to the prejudiced views of customers. If religious clothing causes tensions and disturbances at work, clearly it would be more appropriate to sack the prejudiced employee rather than her religious colleague.

V. Conclusion
This article makes two contributions. First, it identifies several lessons that recent CJEU case law on the protected grounds of religion and belief hold for EU non-discrimination law more generally. The article explains why some groups are protected against discrimination and others are not, and that there should be no differentiation between the groups that are protected. It also seeks to clarify how these groups should be protected-by what standards the boundary between direct and indirect discrimination must (not) be drawn and what conditions a given measure must fulfill to be consistent with the prohibition of discrimination. Second, it offers guidance to legal practice, in particular with regard to the resolution of pending and possible future legal disputes concerning religious equality within the EU. While everyone seems to agree that Achbita, IX v. Wabe, and L.F. v. SCRL-and to a lesser extent Bougnaoui-are missing the point of nondiscrimination law, it has been much harder to identify why exactly these judgments are 163 See Achbita, supra note 1, at para. 42. 164 See id. at para. 43. 165 See Weiler, supra note 5, at 1008. 166 For this argument, see IX v. Wabe, supra note 2, at para. 75. 167 See Shadow Opinion of former AG Sharpston, supra note 9, at para. 125.
problematic. It turns out that the problem is not the definition of direct and indirect discrimination, but rather that the justificatory burden for measures that indirectly burden religious groups is set too low. In particular, the CJEU was too lenient in its assessment of the legitimacy of the measures. Addressing this issue will go a long way towards ensuring that the case law offers sufficient protection for religious minorities from unjustified discrimination.