Michael Connolly

A white worker is dismissed from her job because she married a black man,[1] or a bar denies service to a white woman because she is accompanied by a black man.[2] No one is likely to have to think too long over the policy and technical issues raised in these two scenarios. Any lawyer, let alone a layperson, would consider that this is racial discrimination against the white person. Although atypical, these are ‘simple’ examples of what has become known as associative discrimination. The white person can sue for discrimination because the treatment was based on her association with a black person. The principle is not confined to race; it applies to the usual suspect classes, such as age, disability, gender reassignment, religion or belief, sex, and sexual orientation.

But not all cases will be this simple. Consider this, more complex, scenario: 'A Belfast bakery refused to ice a cake with the slogan "Support Gay Marriage"; it was requested by a customer whose sexual orientation was unknown, and intended for a gay rights group and its celebration of anti-homophobia week.'

This occurred, of course, in Lee v Ashers Bakery (2018). It may surprise some that the UK Supreme Court, reversing all below, utilised the ‘associative’ theory to limit the reach of the UK’s discrimination law, which underpinned the decision that the bakery had not discriminated because of sexual orientation.

In Part 1 of this two-part paper, ‘The ‘associative’ discrimination fiction: part 1’ (published in Northern Ireland Legal Quarterly 72(1) 29–60), I argue that any notion of associative discrimination as a term of art is a fiction and that the Supreme Court’s reasoning was (mis)led by the associative theory into a convoluted and overly restrictive definition of discrimination. The core error was treating the notion of associative discrimination as a term of art, when it has no more utility than providing our vocabulary with a convenient description of a certain factual scenario. Accordingly, in arguing that associative discrimination as a term of art is a fiction, part 1 highlights the likely missteps when treated as such. It does so using this more complex scenario and the reasoning deployed by the Supreme Court to decide the matter.

My analysis of the Supreme Court’s judgment shows that recourse to the legislative provisions would produce much simpler questions to be asked, which were whether there was less favourable treatment because of (or on grounds of) sexual orientation, and if so, did such a finding violate the bakery’s human rights.

The article includes a theoretical overview of the notion of associative discrimination followed by an appreciation of the governing statutory regimes (including those for Northern Ireland and the Equality Act 2010) and general principles of discrimination. The substance of the discussion centres on the reasoning deployed in Lee v Ashers. The court’s associative theory as set out is examined in detail, followed by the other discernible bases of the decision.

These comprise the ‘message case’ exemption and a refusal ‘applying to all’. Both of these bases, it is argued, are unsustainable. The first seems to exempt free speech (including a refusal to say something) from discrimination law. The second, ‘applying to all’, was used to couch the refusal of something facially neutral and therefore not directly discriminatory. Yet, no attempt was made to analyse the facially neutral practice as indirectly discriminatory, which would be the normal course of enquiry.

In part 2 (‘The myth of associative discrimination and the Court of Justice’s great vanishing act: part 2’ (2021) 72(AD1) Northern Ireland Legal Quarterly 1–33, published on 29 July), I contend that the vulnerability of the associative theory for manipulation can work the other way, producing, first, an ‘extended’ notion of associative discrimination and, second, radically broad notions of direct and indirect discrimination. I also argue that a case of the European Court of Justice, CHEZ (2016), heralded as one of associative discrimination, was no such thing. It was a classic Easy Case Making Bad Law. In what was taken wrongly as a novel challenge originating from Bulgaria, a non-Roma resident in a predominantly Roma district claimed that the inconvenience caused by the raising of electricity meters (to prevent tampering) amounted to racial discrimination against her. Both the Advocate General and the court indicated that she should succeed. The Advocate General did so by producing a novel theory of ‘extended associative indirect discrimination’. The court ignored this and, disregarding the traditional form-based technicalities, compared the treatment of the predominantly Roma district with a non-Roma district, and concluded that the treatment of the claimant was direct (racial) discrimination. In other words, an ‘apparently neutral practice’ (the definition given in the Race Directive) was converted from indirect discrimination to direct discrimination, seemingly because of the defendant supplier’s stereotyping of the Roma residents. In straining to find a remedy for the claimant, the Advocate General and the court delivered a morass of novel and convoluted theories of discrimination that will, if nothing else, blur the traditional form-based distinction between direct and indirect discrimination, around which European and UK equality legislation is structured. This result is all the more startling because the obvious solution was provided in the Race Directive’s procedural provisions (article 7), which permits enforcement by persons who consider themselves ‘wronged’ by discrimination. Given that this was clearly discrimination against Roma residents, the solution for the claimant, who was similarly inconvenienced, was straightforward.

[1] See eg Lord Simon, obiter, Race Relations Board v Applin [1975] 2 AC 259 (HL) 289-290 (on the premise that foster parents discriminated against a local authority for refusing to foster children of colour).

[2] See the suggestion by James Comyn QC, approved by Lord Denning MR: Applin v Race Relations Board [1973] QB 815 (CA) 828, and 831 (Stephenson LJ), affirmed [1975] 2 AC 259 (HL).