The right to religious freedom and the right against religious discrimination should be seen as two distinct human rights, with different normative purposes. I have argued elsewhere—with Dr. Jane Norton—that religious freedom is best understood as protecting our interest in our decisional autonomy in matters of religious adherence (and non-adherence). The right against religious discrimination, on the other hand, is a sub-species of the general right against discrimination on protected grounds, and is best understood as protecting our interest in the unsaddled membership of our religious group (akin to our race or tribe). The content of our interest in religious adherence can only be appreciated from the adherent’s internal—committed—point of view; our social, material, and political interest in not being saddled by the (real or perceived) membership of a religious group can be grasped entirely from an external—sociological—perspective.
Dr. Norton and I further explain that the following key doctrinal implications follow from this normative distinction between the two distinct religious interests:
(i) the scope of the prohibition on religious discrimination is narrower than that of the protection of religious freedom,
(ii) the prohibition on religious discrimination may be legitimately extended to certain non-state actors, but the duty to uphold religious freedom should be restricted to the state and state-like bodies,
(iii) establishing a state religion (and other non-zero-sum benefits to a particular religion) may be compatible with the right to religious freedom, but is always discriminatory on the ground of religion, and
(iv) assuming no other right is entailed, it would normally be easier to justify an interference with religious freedom than to justify religious discrimination.
Some of these claims can be clarified by applying our analytic and normative framework to the following hypothetical case: T is a member of the Russian Orthodox Church in Belarus. For most of her life, national identity documents did not include an assigned number as a personal identifier. In 1996, a new form of identity document was introduced that assigned a personal identification number to each citizen. This new form of identity was not motivated by a desire to target or adversely affect any religious group. T refused to obtain such a document, on the ground that her beliefs as an Orthodox Christian prohibited the substitution of a number for a name as the basis for interpersonal interaction, treating humans created by God as soulless objects. T is not alone in asserting this type of objection, which is shared by certain other Orthodox Christians, certain members of other Christian denominations, and certain members of non-Christian religions. However, the Holy Synod of the Russian Orthodox Church issued a statement in 2001, explaining that numbers on personal identity documents have no religious significance. As a result of T’s refusal to obtain a national identity document in the new required form, the government of Belarus denied her application for certain government-supplied benefits to which she was otherwise legally entitled. The government refused to accept any alternative form of proof of her identity, and relied on its consistent practice of requiring current national identity documents from all applicants for such benefits. She challenges the denial of these government benefits as a disproportionate restriction of the right to freely practice her religion, and as indirectly discriminating against her on the basis of her religious beliefs.
On T’s religious freedom claim:
- Since the interference is through state action, T should have the right to bring a claim. Had the facts been different, and a number-based identification was used by a small private employer, T should not have been able to sue. Religious freedom should not be enforced horizontally between X and Y: just as X has the freedom to adhere to her religion, Y should also have the freedom to not be obliged to do anything (or refrain from doing something) because of X’s religious adherence. That would be an illegitimate interference with Y’s freedom from X’s religion.
- The relevant interest engaged with respect to T’s religious freedom claim is her interest in freely adhering to the beliefs and practices of her religion. The scope of what T’s religion demands of her can be properly assessed only from her own committed, internal, viewpoint. Given the remarkable diversity and plurality in religious adherence, both between religions and within them, it would be a severe curtailment of religious freedom of dissenters and heterodox believers—incidentally, those most in need of protection—to suppose that only the “official doctrine” of that religion deserves the law’s protection. Very little will be left of religious freedom if the state is allowed to tell adherents that their religion is not, in fact, what they think it is. Therefore, the Holy Synod’s view on the matter is irrelevant. Also irrelevant is any state official’s own interpretation of the demands of Russian Orthodox Christianity on T. Furthermore, whether T alone has this interpretation of her religious tradition, or shares it with several others, is also beside the point. So long as T’s claim is sincerely made, and the object of her claim—personal identification by a number—has a plausibly religious character, T’s interest in her autonomous exercise of religious adherence is engaged.
- T is likely to satisfy the sincerity and plausibility requirements, which are relatively low-threshold gatekeepers of the right to religious freedom. Thus, a number-based personal identifier is a clear interference with T’s interest in her religious adherence.
- The outcome of the case will turn on whether the interference is justified. This will depend on the outcome of a proportionality-based justification analysis that weighs and balances T’s interests against those of the state and any third parties. A more granular access to the factual context is required to say whether the interference is proportionate, and therefore justified. The following factors should be relevant:
- Interference with religious freedom often implicates the adherent’s other fundamental freedoms. Burqa bans, for example, engage her freedom of expression, identification photographs engage privacy, and so on. It does not seem, at least prima facie, that any of T’s other fundamental freedoms are engaged in this case. There is no restriction on her freedom of expression, movement, association, privacy, occupation etc, or, at least, no more restriction than a document that identifies her by name. Her objection is not to an identity document itself, but one that identifies her by a number. So, her religious freedom claim in this case is not buttressed by other fundamental freedoms, and is, therefore, relatively weaker.
- A court should consider the legitimacy and the weight of the state’s interest in numerical identification. It may be that numerical identification is more reliable and efficient, or that a name-only identification system is no longer feasible in a large population, or that it is unacceptable to other jurisdictions whose recognition for this document is essential (especially if it is a passport).
- The consequences of T’s exercise of her religion also matter. Here, the impact is serious—she is denied state benefits, which she is presumably otherwise entitled to. The state will also need to show that there is no alternative identification system it can feasibly rely upon to check her eligibility for these benefits. The case will turn not only on the state’s reasons for preferring identification by numbers, but also the costs of either retaining the old system for everyone, or at least for those who oppose numerical identification for religious reasons, or accepting alternative modes of identity verification at least in some contexts. If these costs are relatively small, or an alternative system that does not implicate religious freedom can adequately serve the state’s objectives, T’s claim should prevail. If not, she should fail.
On T’s religious discrimination claim:
- There is no direct discrimination against T on the basis of her religion because the rule in question is facially neutral and (we have assumed) not designed as a pretext to target any religious group. The only issue is whether she has been discriminated against indirectly.
- Prima facie indirect discrimination occurs when a facially neutral rule, practice, or policy has a disproportionate impact on a protected group.
- For the purposes of indirect discrimination law, the “group” in question is understood loosely as a collection of people sharing the protected characteristic—it need not (but sometimes will) be a close-knit and cohesive community sharing a history, ethnic origins, culture, language etc. For the purposes of discrimination law, for example, disabled persons constitute a “group,” whether or not they share a sense of belonging.
- The group, however, must be a sociological entity, made up of those who share the protected characteristic (in this case, members of T’s religious group). It cannot be a subjectively imagined collection of individuals. Unlike T’s religious beliefs, which she is largely free to determine subjectively, which religious group she belongs to is determined not from her internal point of view but from the external, sociological, point of view. A religious group, in this sense, includes people who are or are taken to be members of that group (because they share its dress code, culinary habits, etc). Atheistic Muslims who have moved to England from Bangladesh, for example, may not be adherents of Islam, but their group membership is still likely to be Muslim—and they will continue to share the burdens and privileges of this membership—irrespective of the fact that they have renounced religion. Cultural Christians and many recent converts also see a divergence between their religious adherence and their religious group membership. In T’s case, let us assume that there is no divergence between her adherence and her group membership. If so, the relevant religious group she longs to is likely to be the Russian Orthodox Christian community in Belarus (some of whom may have converted to other religions or even given up on religion entirely; many of whom, even if co-believers, may not share T’s particular objection to numerical identification).
- T’s “religious group” cannot be all those Belarussians, or even all those Russian Orthodox Christian Belarussians, who share her belief that a numerical identification breaches the tenets of Russian Orthodox Christianity. That way of characterising a group ignores the conceptual distinction between T’s interest in her religious adherence and her interest in not being saddled by her religious group membership, as well as the normative distinction between the two rights that track these interests. Constituting her religious group in terms of this specific belief would not only collapse the distinction between religious discrimination law and religious freedom law, it would also end up characterising every case of indirect religious discrimination as a case of direct religious discrimination, by definitional fiat.
- So, the empirical question before the court would be whether the new identification document’s numerical basis for identifying citizens disproportionately burdens the members of the “religious group” that T belongs to. Here, there are three possibilities, since T is simultaneously a member of three sociological groups defined by her religious membership:
- Russian Orthodox Christians in Belarus
- Christians in Belarus
- Religious adherents in Belarus
- In order to find prima facie indirect discrimination, T will need to show that the group of people adversely affected by the numerical identification policy (let us call this set of persons “V”) is constituted disproportionately by members of one of the three aforementioned groups. This is necessarily a comparative analysis, in relation to those who are not members of the respective group in each case.
- Since the issue involves identification documents for all citizens, the difficult question of the relevant pool for identifying the set V is happily avoided in this case. That issue arises most controversially in employment cases—since discrimination law governs pre-contractual as well as post-contractual relationships, it is not clear whether V should be drawn only from existing employees, from employees and job applicants, or from the relevant labour market. In such cases, choosing a larger pool is often fairer to the claimant doctrinally, but harder to prove evidentiarily.
- In statistical terms, a protected group is adversely affected disproportionately if a person’s membership of that group is a significant predictor for her belonging to the set V. Claimants have typically found satisfying this technical standard extremely difficult in actual cases. For pragmatic reasons, therefore, many jurisdictions no longer require statistical proof of disproportionality, and are satisfied by rule-of-thumb evidence.
- The assumed facts are not sufficient for us to know whether even this rule-of-thumb standard of disproportionality will be satisfied vis-à-vis her membership of any of the three groups. If the case concerned a ban on the wearing of all face-coverings in public, for instance, it should be clear that a Muslim claimant’s likelihood of showing disproportionality would have been higher with regard to her membership of Belarussian Muslims compared to her membership of the group comprising all religious adherents in Belarus. This intuition is based on some background empirical assumptions about the practice of face covering in different religious groups. In the numerical identification case, I am unable to say whether such a relative judgment about the likelihood of success by invoking T’s membership of different religious groups can be made.
- Assuming that T does succeed in showing disproportional impact based on her membership of at least one of the three groups, and thereby establishes prima facie indirect discrimination, the state will need to justify its numerical identification policy. It will have to show that the disproportionate adverse impact on T’s group is a suitable, necessary, and proportionate means of pursuing a legitimate state objective. Although the test applied in this case—that of proportionality—is the same as the one used to decide whether Belarus’s interference with T’s religious freedom was justified, the manner in which the test will apply here is different.
- In general, it will be harder for the state to justify prima facie indirect discrimination than to justify an interference with religious freedom (assuming no other fundamental freedom is engaged in either case). This is because of the following factors:
- While the general interest to freely adhere to one’s religion has substantial weight, secular courts cannot determine the importance of a particular religious belief or practice to the claimant when adjudicating religious freedom claims. They must, of necessity, assume that all such claims have a significant weight, but that weight cannot be deemed to be so insurmountable that governance by general laws becomes difficult. Governance would become next to impossible if we assume a substantial weight for all religious claims alongside the necessarily broad and subjective scope of that right that I suggest flows from the internal nature of religious adherence. In other words, a shallow bite is an inevitable consequence of the broad scope of the right to religious freedom. This is assuming that the infringement of religious freedom is a side-effect of a general law, and is not designed for that purpose.
- The interest in unsaddled membership of one’s (chosen or unchosen) religious group, on the other hand, is usually weightier. Substantial advantage gaps between different religious groups create a second-class status for (actual or perceived) members of disadvantaged groups, compromising their ability to access key basic goods.
- The particular external consequences for the claimant that flow from the infringement of either right is publicly verifiable for both rights. In this case, it is the inability to access state benefits, which surely has significant weight. Even so, if this benefit is being denied to a protected group (that is, a group defined by normatively irrelevant characteristics such as its real or perceived race, religion, sexual orientation etc), such denial may reinforce or exacerbate its lower social, political, or material status. The harm of denying a benefit to an individual is bad enough, its denial to a vulnerable group especially odious. Assuming that access to the benefit in question is itself not a right, the law is likely to demand a higher justification threshold for the latter type of denial, and will be right in doing so.
* Professor of Law at the Universities of Oxford and Melbourne, and Global Visiting Professor at New York University Law School.
 This claim is developed and defended in T Khaitan and J Norton, “The Right to Freedom of Religion and the Right Against Religious Discrimination: Theoretical Distinctions” 17(4) International Journal of Constitutional Law (2019) 1125. See also, T Khaitan, “The Point of Discrimination Law” in Nussbaum et al, The Empire of Disgust (OUP 2018).
 We track these implications in T Khaitan and J Norton, “Religion in Human Rights Law: A normative Restatement” 18(1) International Journal of Constitutional Law (forthcoming 2020), available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3274124.
 Bills of Rights (in jurisdictions other than the United States) tend to frame rights in a qualified rather than an absolute manner (exceptions include the right against torture, which is always absolute). Such qualified constitutional rights are typically applied through a two-stage analysis: the first stage determines whether there has been a prima facie infringement of the right; if yes, then at the second stage, courts ask whether such infringement is justified, usually by asking whether the infringement is “reasonable” or “proportionate.” See generally, T Khaitan, “Beyond Reasonableness—A Rigorous Standard of Review for Article 15 Infringement” (2008) 50(2) Journal of the Indian Law Institute 177–208.
 Based on Yachnik v Belarus (Decision of the Human Rights Committee in 2014 in Communication Number 1990/2010). I am grateful to Prof Gerald Neuman for providing this helpful adaptation of the facts of the case.
 For details of how the plausibility and sincerity requirements are worked out and applied, see Khaitan and Norton, supra note 1 and 2.
 A proportionality analysis is extremely familiar to many jurisdictions in Europe, and in Canada and South Africa. Although its minutiae remain controversial, broadly speaking, it is a means-end analysis that examines the legitimacy of the state’s end, the suitability and the necessity of the means adopted, and a cost-benefit balancing of the trade-off between the end and the right in question.
 T Khaitan, “Indirect Discrimination” in K Lippert-Rasmussen ed., The Routledge Handbook of the Ethics of Discrimination (Routledge 2018) p 32.
 T Khaitan, A Theory of Discrimination Law (OUP 2015), pp 30, 51.
 Khaitan and Norton (2019) p 1134.
 Khaitan and Norton (forthcoming 2020), final paragraph under section 2.
 This would be the consequence at least in British and European law, where intention is not a necessary ingredient to prove direct discrimination, so long as the claimant can show an exact correspondence between those who are adversely affected and not affected on the one hand and a protected group and its cognate on the other. See generally, Khaitan, in Lipprt Rasmussen ed (2018) p 36.
 Khaitan (2015), p 155.
 H Collins and T Khaitan, Foundations of Indirect Discrimination Law (Bloomsbury 2018), p 14.
 T Khaitan and S Steel, “Wrongs, Group Disadvantage, and the Legitimacy of Indirect Discrimination Law”, in Collins and Khaitan (2018), p 203.
 Khaitan and Norton (forthcoming 2020), text to footnotes 37-41.
 Khaitan, in Nussbaum ed, (2018).