: are UK claimants at a disadvantage?

The minimum standard of scrutiny for the justification defence in the context of indirect discrimination was first set out by the Court of Justice of the European Union (CJEU) in Bilka-Kaufhaus GmbH v Weber von Hartz (1986). This established that an indirectly discriminatory measure is justified if it meets a real need and is appropriate and necessary for meeting that need. The UK courts’ approach to the concept of proportionality within the context of this justification defence may nevertheless have distinct disadvantages for claimants in comparison with their EU counterparts. The approach of the UK courts is assessed here by considering the development of case law in this area, both in the Employment Appeal Tribunal and in the higher courts. When compared to the approach taken by the CJEU, it becomes apparent that there is a significant difference between the ways in which UK courts and the CJEU interpret the justification defence. Findings show that the approach of the UK courts significantly disadvantages claimants, leading to the conclusion that the UK may not be fully compliant with EU law. To remedy this defect, it is suggested that there are at least two practical alternative solutions. The first is that Parliament could incorporate a strict necessity test into the Equality Act. Alternatively, the courts could develop a ‘robust approach’ to proportionality. The outcomes of a large number of employment law cases are examined here, appearing to suggest that the latter approach may have greater benefits for claimants than those associated with adopting a strict necessity test, although it is unlikely that will find favour with either Parliament or the courts.


INTRODUCTION
The Equality Act 2010 (EqA) makes discrimination in the workplace towards protected groups unlawful. Indirect discrimination is not unlawful, however, if it is justified as being 'a proportionate means of achieving a legitimate aim' . 1 In assessing the merits of justification, courts and tribunals must apply the principle of proportionality, with the aim of striking a balance between the interests of the employee and the employer.
The interpretation of justification is crucial to determining the extent of protection for UK claimants from indirect discrimination. This article compares the interpretation given to justification by the domestic courts against those of the Court of Justice of the European Union (CJEU). It evaluates the concern that the approach of UK courts and tribunals is less rigorous than that of the CJEU, to the detriment to UK claimants.
These claims are assessed by reference to empirical evidence from a number of Employment Appeal Tribunal (EAT) decisions, to discover whether by applying different approaches to the assessment of justification in these cases, the outcome of the case would thereby be altered. It is also suggested that a way of increasing protection for claimants, based on existing UK case law, may be to render the proportionality assessment more robust by requiring that further enquiries are routinely made in order to ensure that there is sufficient evidence as to the discriminatory impact of a measure.

JUSTIFICATION: UK and EU LAW
There are significant differences in the approach to justification taken by the UK courts in comparison with the CJEU. Despite initially setting a high threshold for justification in Steel v Union of Post Office Worker (1978), 2 where justification was equated with business necessity, in Ojutiku v Manpower Services Commission (1982) Eveleigh LJ defined justification as 'acceptable to right-thinking people as sound and tolerable reasons', 3 thereby equating justification with the far less rigorous standard of reasonableness.
Nevertheless, in Bilka-Kaufhaus v Weber von Hartz (1986), the CJEU determined that indirectly discriminatory measures could be justified only if they 'correspond to a real need on the part of the undertaking, are appropriate with a view to achieving the objective in question and are necessary to that end ' . 4 This equated justification with necessity-a much higher standard than that applied in Ojutiku, and more closely allied to the test in Steel.
In an attempt to clarify the issue, the Court of Appeal (CA) again considered what the correct approach to justification should be in Hampson v Department of Education and Science (1986). 5 The CA, however, avoided imposing a strict necessity requirement, as set out in Bilka, by advocating a proportionality assessment which balanced the discriminatory effect of a provision with the reasonable needs of a business. This approach was subsequently approved by the House of Lords in Webb v EMO Air Cargo (UK) Ltd (1993), 6 thus settling the direction taken in UK case law in this area ever since.
These developments in case law, where the question has been whether or not indirect discrimination was justified, have allowed UK courts and tribunals considerable scope when considering the correct definition of justification. It has enabled them to avoid applying a strict test of necessity and to retain the familiar common law concept of reasonableness in the form of reasonable necessity. This absence of a strict necessity test has led equality campaigners and politicians to express concern that UK claimants may be afforded less protection from indirect discrimination than that intended in EU law. 7 These concerns have also been expressed by several academics. Hervey, writing shortly after the decision in Hampson, suggested that UK courts were not following the principles laid down by CJEU in relation to justification. 8 Connolly expressed similar concerns, demonstrating by reference to high profile cases such as Enderby v Frenchay Health Authority 9 how the absence of a strict necessity test in domestic law gave employers 'more leeway' to discriminate against employees. 10 More recently, shortly before the enactment of the EqA, Baker expressed concern as to the relative weakness of the UK standard of justification as compared to that of the CJEU. 11 Baker also expressed the view that over 20 years of UK jurisprudence on the matter made it unlikely that the courts would 'back pedal' and adopt a test of necessity in relation to justification, but he did suggest that it was possible to develop a more robust model of proportionality, based on existing case law, which could potentially strengthen the protection from indirect discrimination, 12 a possibility which is given further consideration below.
This article builds on the work of these earlier commentators by examining, with reference to empirical evidence, whether the difference in approach by the UK courts results in UK claimants being treated less fairly because of the failure to impose a strict necessity test in relation to indirect discrimination.

PROPORTIONALITY: CASE LAW
Although the evidence suggests that UK claimants may be at a disadvantage, a proportionality assessment has the potential to provide more effective protection for claimants. A robust proportionality assessment could even deliver greater protection than that provided by a strict test of necessity, since a significant discriminatory impact could outweigh even necessity. In cases where the discriminatory impact of a measure is fully considered, the intensity of scrutiny afforded by a proportionality assessment is considerable. 13  of direct discrimination, most of the Supreme Court judges also considered the issue of indirect discrimination and justification. What stands out from a reading of the judgments is that the outcome with regard to justification differed according to the individual judge's application of proportionality.
The facts of the case were that a child was refused admission to an Orthodox Jewish school on the grounds that his Jewish heritage had not descended through the maternal line. Lord Mance adopted a broad approach to the question of justification. This included consideration as to whether the school had established that other less discriminatory alternatives were unsuitable to achieving a legitimate aim, but he went even further than this by considering the impact on society of maintaining the discriminatory policy. This led Lord Mance to the conclusion, with which Lady Hale, Lord Kerr, Lord Hope and Lord Clark concurred, that the policy was not justified. 16 In contrast to this broad approach, Lord Roger and Lord Brown took a much narrower view of what proportionality means, giving greater consideration to the needs of the discriminator, with much less emphasis on discriminatory impact. On this analysis, the policy was justified. 17 JFS and Elias illustrate how the intensity of review in proportionality assessments is substantially increased if adequate consideration is given to discriminatory impact. Unfortunately, this particular evidence suggests that the approach outlined in these cases is rarely applied at the level of the EAT. This applies to some of the more recent cases that are considered below. 18 Furthermore, there are other precedents in which the importance of considering the discriminatory impact in its fullest sense is given considerably less emphasis. For example, in Barry v Midland Bank, 19 a case which is still frequently cited and followed in EAT decisions, the discriminatory impact of a voluntary severance scheme which was alleged to have discriminated against Mrs Barry as a part-time worker was not fully examined, the justification issue not even given consideration. If tribunals continue to have a choice of approaches in relation to proportionality-and the evidence in this study suggests that the approach in Barry is more commonly applied than the approach in Elias and JFS-this can only serve to exacerbate the problem of claimants being treated unfairly compared with those whose claims reach the CJEU.
Overall, courts and tribunals seem to be far more concerned with investigating the needs of employers than considering the discriminatory impact of a measure in its fullest sense. An illustration of this imbalance is Pill LJ's comments in Hardy & Hanson plc v Lax 20 where he stated that a 'broader understanding of the needs of business will be required' in cases concerning justification. 21 Pill LJ did not add that tribunals also had a proactive duty to fully consider discriminatory impact within the proportionality assessment. It is not surprising, therefore, that the EAT appears to give this aspect of the proportionality assessment scant consideration. Furthermore, in cases where the EAT fails to fully consider discriminatory impact, this disadvantage will not be offset by a strict approach to necessity as it is still the view of the Supreme Court that the correct test is reasonable, rather than strict, necessity. Lady Hale confirmed in Homer v Chief Constable of West Yorkshire Police [2010] 22 that reasonable necessity is the correct standard to be applied in relation to justification.

A. Comparing Different Approaches to Justification at the EAT
Given the lack of a strict necessity test, as per Bilka, it is at least questionable whether the UK is fully compliant with EU law in relation to justification. Recent case law coming from the CJEU has confirmed that the Bilka test also applies to the new categories of protected groups outlined in the Equality Directive. 23 UK law should, therefore, be providing equivalent protection to that provided by the Bilka criteria to all protected groups. We sought to examine whether this is, in fact, the case by considering a number of EAT decisions made between 1992 and 2015. We also sought to establish whether any deficiency in the UK approach could be remedied by applying a more robust approach to proportionality. 20  We considered 44 EAT cases which were heard during this period and which involved issues of indirect discrimination, proportionality and justification, 24 analysing the decisions made in those cases in relation to the following: (i) The general approach of the EAT in relation to proportionality; (ii) Whether applying the Bilka test of justification made a difference to the result in individual cases; (iii) Whether applying a robust approach to proportionality, whereby discriminatory impacts were weighed in the fullest sense, made a difference to the result in individual cases.

B. The Successful Cases
Of the 44 cases, 21 cases 25 resulted in a positive outcome for the claimant in that justification for the alleged discrimination was not established by the employer, while 23 claimants were unsuccessful. 26 Of the successful cases, eleven resulted in the overturning of the decision of the Employment Tribunal (ET) on the issue of justification, while ten confirmed the decision of the ET. Although five of these cases were appealed further, none were overturned on the issue of justification. 27 Further analysis of the successful decisions reveals a high level of consistency in the application of the Hampson test of proportionality, in which discriminatory impact is weighed against the reasonable needs of the employer. This approach was applied in most of the cases analysed, although some equal pay cases did confine themselves to applying the jurisprudence of the CJEU. 28 It is disturbing to note that some of the cases in which the ET decision was overturned suggest an insufficient understanding of the relevant legal test for justification at tribunal level. In Games v University of Kent (2014), for example, a worker who was near to retirement was required to obtain a PhD in order to continue working as a lecturer. 29 This measure indirectly discriminated against the claimant on the grounds of age. The ET determined that the discrimination was justified as, prior to the provision being applied, the claimant had been given ample opportunity to obtain the qualification. The acceptance by the ET of such spurious justification indicates a substantial misunderstanding of the relevant legal test.
Other Although the employer was unsuccessful in establishing justification, it is important to note that the employer perceived the test of reasonable necessity to be less onerous than a test of strict necessity. This supports the assertion that one danger of adopting a test of reasonable necessity is that employers may perceive that this provides some leeway with regard to justification, whereas a test of strict necessity makes clear that only the least discriminatory alternative will be justifiable. 31 When examining the decisions where the claimant was successful, it soon becomes evident that in some areas the EAT is applying a more robust level of scrutiny to the issue of justification. This is particularly noticeable in equal pay cases. In Davies v Neath Port Talbot CBC 32 the EAT overturned the decision of the ET that the employer was justified in not paying fulltime wages to a trade union representative while attending a full-time trade union course as the claimant normally worked part-time. The EAT directly applied the jurisprudence of the CJEU in Arbeiterwohlfahrt der Stadt Berlin v Botel, 33 finding that there was no justification for failing to compensate the applicant in full.
In fact, our analysis reveals a closer adherence to EU law in equal pay cases than in cases concerning other types of discrimination, with the majority of such cases resulting in a successful outcome for the claimants. This may be due to the wealth of EU jurisprudence in this area, but may equally be attributable to the fact that many of the equal pay cases concerned multiple claimants represented by large organisations such as trade unions with the resources to appeal these cases further, thereby ensuring effective representation and full consideration of the issue of justification. It is also possible that tribunals have become well informed as to the considerable discriminatory impact caused by pay differentials and automatically accord this type of discrimination a high weight within the proportionality balance.
All the cases which were successful for the claimant would also have been successful if the Bilka test had been applied and all but one would have 30 South Tyneside Metropolitan Council v Anderson [2007] EWCA Civ 654; [2007]  been successful if a robust application of proportionality had been adopted. The case in which a robust application of proportionality might have produced a different result from the Bilka test is Redfearn v Serco. 34 Redfearn was subsequently appealed to the CA and the European Court of Human Rights (ECtHR), 35 but it is the decisions at the level of the ET and EAT that we are concerned with here. The facts in Redfearn were that the claimant became a councillor for the British National Party (BNP) which led to his dismissal from work. Mr Redfearn took a claim of racial discrimination to the ET. The ET considered, inadequately, the issue of indirect discrimination, deeming the employer's actions to be proportionate to achieving the legitimate aim of maintaining health and safety, as the claimant's work necessitated frequent contact with ethnic minorities to whom the BNP was known to be hostile. On appeal, the EAT determined that the ET had provided inadequate reasons for reaching this decision, taking into account that the dismissal had a severe impact on the claimant and that other options were available to the employer which could have avoided the claimant having to work directly with the public. If the ET had applied the Bilka approach to this case, the existence of suitable, less discriminatory alternatives would almost certainly have defeated the employer's defence of justification. If, however, a wide consideration of discriminatory impact had been fully considered by the ET, the employer might have succeeded in establishing justification. This is because the impact on society as a whole could have been considered. This would have involved weighing up the consequences of the infringement on the claimant and the full impact of any discrimination against the needs of the employer and the impact on society of the discriminatory measure. As this overall picture would have included the adverse effects on society of promoting politics which endorse discrimination, it is likely that the employer's need would have outweighed-and therefore justified-any discrimination in this case.
Redfearn v Serco is a rare example of how a wide consideration of the discriminatory impact could potentially produce a beneficial result for the discriminator and which promotes the spirit and purpose of the equality legislation. This wider consideration reflects a purposive approach, resulting in a decision that is more in accordance with the intentions of anti-discrimination legislation, the intention of which cannot have been to protect those who promote views that endorse discrimination. An analysis of this kind 34  might also have prevented the case being appealed further, as some of the issues which were subsequently successful on appeal would have been considered under a wider application of proportionality.

C. The Unsuccessful Cases
In 23 of the cases which we examined, the claimant was unsuccessful in that the EAT deemed the discrimination to be justified. It is our contention that applying the Bilka test of strict necessity could have led to a different result in relation to justification in 13 of these cases, while applying a robust approach to proportionality would have led to a different result in 12 of them. 36 In theory, therefore, the Bilka test of strict necessity which is applied by the CJEU could produce a more favourable result for claimants than the Hampson test does. The difference in result between the application of the Bilka test and the application of a robust approach to proportionality is, however, much less certain. In terms of explaining these differences, it is necessary to look in detail at some of these cases.
In Burch v Tesco Stores Ltd, 37 concerning sex discrimination, a personnel manager was denied promotion on the grounds that she would struggle to work more than 32 hours a week due to the demands of her own childcare responsibilities. Tesco claimed that it was a necessity for senior management to be flexible and available for work 24 hours a day. The ET found the discrimination, in this case, to be justified by the employer's need. The EAT allowed this appeal on other grounds, but confirmed the ET's decision on justification, despite the fact that there were alternatives which would have met the employer's need for flexible cover with a lower discriminatory impact. The EAT stated: The civil law of Europe would no doubt describe it as applying the principle of proportionality. The balancing exercise by its phraseology is not a question of considering absolutes it is a matter of balance and none of the language shown allows extremes to be urged on one side or the other, just as justification is no more than of convenience, so need does not mean necessity. 38 This suggests that the EAT applied a test of reasonableness rather than proportionality in this case. Furthermore, the balancing exercise as described by the EAT does not acknowledge that a high weight should be accorded to discriminatory impact or that justification is a defence for the employer to prove. If a strict test of necessity had been applied in Burch, 39 the existence of suitable alternatives would almost certainly have led to a finding that the discrimination was not justified.
Alternatively, if a robust test of proportionality had been applied, with a wide consideration of discriminatory impact, the EAT would have considered the effect on society of excluding employees with childcare responsibilities (who are predominantly women) from senior management positions. The EAT records that at the time of the hearing Tesco employed over 167,000 staff 40 which severely weakens the claim that the employer had a real business need for such flexibility from an individual member of staff. This need would, therefore, hang lightly in the proportionality balance.
A further case which might have been decided differently under the Bilka test is Azmi v Kirklees Metropolitan Council, 41 in which a devout Muslim teaching assistant was dismissed for refusing to remove a veil which obscured her face. The veil had been found to hamper her ability to make herself heard when teaching. The employer could, therefore, demonstrate a real need for the claimant to remove her veil in these circumstances.
The claimant agreed to an adjustment by which she would remove her veil when male staff were not present. This necessitated a move to another year group which consisted of predominantly female staff members. The school refused to allow this as the claimant worked part-time and the school preferred full-time workers to work with younger year groups. No evidence, however, was presented to demonstrate that this constituted a real need on the part of the employer. The EAT applied a test of reasonable need and found the discrimination against the claimant to be justified. If, however, a strict test of necessity had been applied, the alternative of allowing the claimant to work with another year group would have been more 38  thoroughly investigated. If this proved to be a suitable alternative, the discrimination would not have been justified.
If a robust application of proportionality had been applied, the EAT would have had to consider the effect of discrimination against the claimant in a wider sense. This would have included the effect on the predominantly Muslim local community, the number of Muslim women who might be affected by the provision, and the overall societal impact. A recent study clearly demonstrates the societal impact of discrimination against Muslims, finding that Muslims face the worst employment discrimination of any minority group in the UK. 42 This significant discriminatory impact should, therefore, be an important consideration in claims of indirect discrimination by Muslims.
In another case involving a Muslim worker, Cherfi v G4S, 43 the claimant worked as a security guard for a large firm. For several years, the claimant left work for an hour on Friday lunchtime to attend prayers at a local mosque. The claimant's employer entered into a contract with Jobcentre-Plus whereby security guards were required to be in attendance throughout their shift, including the lunch hour, meaning that the claimant was no longer able to attend Friday prayers. This caused considerable difficulties for the claimant who was then unable to work on a Friday. The ET which considered the case found that there was discrimination in this case but that it was justified by the employer's need to fulfil the terms of their contract. The EAT accepted that the ET had carried out the necessary balancing exercise, weighing the employer's reasonable needs with the discrimination against the claimant, and confirmed the ET's decision. As in the previous case, however, very little weight was given to discriminatory impact. In fact, consideration of discrimination was limited to the effect on the individual worker. Furthermore, little attention was given to the possibility of alternatives that would have met the employer's need with less discriminatory impact.
If the Bilka approach had been applied to this case, the employer would have had to demonstrate that it was necessary for the claimant to be on site throughout the day on Friday and that there were no less discriminatory alternatives available to fulfil this need. The employer stated that it was 'not practicable to bring in another guard to cover the claimant's lunchtime absences' . Nevertheless, given the size and resources of this company, it is 42 N. Khattab  suggested that practical alternatives to meet the employer's need should have been more fully explored. Applying the Bilka test, the employer would have had to establish that it was not possible, rather than not practicable, to make an alternative arrangement. A robust application of proportionality would have revealed a considerable discriminatory impact, which was barely considered by the EAT. The societal impact of this discrimination is substantial, since it hinders the employment prospects of a minority group who already suffer significant workplace discrimination. This considerable discriminatory impact would be weighed against the employer's real need to fulfil its contract. As the employer had multiple resources, and there were other less discriminatory alternatives available to fulfil the employer's need, it is likely that if a robust application of proportionality had been applied here, the employer would have been unable to justify the discrimination. It is important to note that Cherfi was decided at a later date to both the CA decision in Elias and the Supreme Court ruling in JFS, suggesting that the approach to proportionality outlined in these decisions is routinely ignored at the level of the EAT.
Cherfi can be contrasted with Mba v London Borough of Merton. 44 In Mba, the claimant was a residential care officer and a devout Christian. She worked in a care home, looking after children with severe disabilities who required round-the-clock care. The claimant's religious beliefs meant that she was unable to work on a Sunday; this need was accommodated for two years, but eventually, it began to cause problems with staffing. This led the employer to decide that the only way to maintain adequate staffing was to include the claimant in the Sunday rota. When she refused to work on Sundays she was disciplined, and she alleged indirect religious discrimination. The ET accepted that she had been discriminated against, but found the discrimination to be justified when applying the Hampson test. The EAT affirmed this position.
Had the Bilka test been applied to Mba, the claimant would probably have succeeded, since there were less discriminatory alternatives available which could have met the employer's need. If, however, a robust application of proportionality is applied to this case, it is not clear that the claimant would have succeeded. This is because the discriminatory impact in a wider sense is relatively low. Indeed, the research conducted by Khattab and Johnston on religious discrimination in the UK workplace found that Christian employees suffered very little workplace discrimination. 45  relation to proportionality, the discriminatory impact is less. This relatively low discriminatory impact could be outweighed, even in a robust application of proportionality, by a real need on the part of the employer.
There is still a requirement, however, for the employer to provide evidence of a real need that outweighs the discriminatory impact. This is illustrated in Eweida v British Airways Plc, in which the claimant, also a devout Christian, worked for British Airways. 46 The claimant wished to wear a small cross on a necklace, but this contravened her employer's dress code. She was disciplined when she insisted on wearing the cross. The ET, EAT and the CA determined the case against her. However, when she took her case to the ECtHR, the issue of justification was decided in her favour. 47 It is interesting to compare this decision with that in Mba. While the discriminatory impact in both Mba and Eweida was relatively low, the need for British Airways to maintain a dress code without religious symbols was also quite weak, leading to a finding that the impact on the claimant outweighed the employer's need. It is suggested that if the Bilka test had been applied to Eweida, the claimant would also have succeeded, as the employers' need for a smart corporate image could have been achieved with less discriminatory impact by making some small adjustments to the dress code (adjustments which British Airways had in fact already implemented by the time of the hearing).
Our analysis of EAT decisions has also revealed that in most cases the ET and the EAT do not apply a strict test of necessity in relation to justification. This is apparent in a recent decision of the EAT, Kapenova v Department of Health. 48  Our analysis of EAT decisions has demonstrated that the absence of a strict test of necessity can, in most circumstances, be compensated for if the proportionality assessment includes a wide-ranging consideration of discriminatory impact. Nevertheless, the evidence suggests that in most cases neither the EAT nor the ET gives this issue sufficient consideration. This is despite the existence of judicial precedents such as JFS and Elias. Although justification is a defence, which the employer has to prove, the EAT appears to attach more importance to obtaining evidence of employers' needs than to consideration of discriminatory impact. Cases such as Cherfi and Burch demonstrate the detrimental effect that this has on claimants where the discriminatory impact is significant but is never properly weighed in the proportionality balance.

RELIGIOUS DISCRIMINATION
Our analysis of EAT decisions provides evidence that those claiming that they have been subject to religious discrimination may be suffering disproportionately. Eight of the 23 unsuccessful cases analysed concerned religious discrimination. This is of particular concern in cases concerning Muslim claimants, as evidence suggests that discrimination against this group is a serious problem. 50 The discriminatory impact in cases concerning Muslims should, therefore, be given considerable emphasis. The evidence of this study demonstrates that this weighty discriminatory impact is inadequately considered by the EAT, and we suggest that it is imperative that courts and tribunals are better informed in this area.
With regard to religious discrimination, concerns have been voiced that, in practice, religion is afforded less protection than that afforded to other protected characteristics. 51 There is evidence from case law which suggests that courts and tribunals may, whether consciously or unconsciously, be engaging in a ranking exercise in relation to protected characteristics, to the detriment of religious belief. This seems particularly prevalent in cases where competing rights are concerned. For instance, in Ladele v Islington Borough Council 52 supporting the 'need' of a Christian registrar to abstain from samesex ceremonies could be seen to endorse discrimination towards same-sex couples. Although we sympathise with the difficulties that tribunals face in these cases, we nevertheless submit that ranking protected characteristics is not an appropriate way to deal with competing rights. Indeed, the case law of the CJEU has clarified that the level of protection afforded to different protected groups should be consistent. 53 Furthermore, denigrating the importance of religious discrimination has the potential to create injustice for religious groups where the discriminatory impact is high. The evidence in this study shows that ranking is unnecessary if proportionality is applied in a robust way, as a low discriminatory impact will always be outweighed by a real need on the part of the discriminator. In addition, competing rights are better dealt with by a wide-ranging consideration of societal impact. 54

THE WAY FORWARD
Our evidence suggests that UK claimants may be losing out in cases of indirect discrimination and that the UK is currently failing to fully comply either with the letter or with the spirit of EU law. This is due to two factors: the failure of UK courts and tribunals to apply a strict test of necessity and the inadequate weight given to discriminatory impact within the proportionality assessment.
Furthermore, the minimum standard of scrutiny for justification has been clearly set out by the CJEU in Bilka and in Article 2(2)(i) of the Equality Directive. 55 The UK does not have to implement the provisions of this Directive in a literal way, but domestic anti-discrimination must, at the very least, produce a result that is as effective as that outlined in the Directive. 56 The evidence examined here, which is both wide-ranging and representative, suggests that the approach of the UK courts is not fully compliant with this standard.
To remedy these shortcomings, it is suggested that there are two possible options. The first option would be for the UK to adopt a strict necessity test in relation to justification. This could be achieved in one of the two ways. First, by amending the EqA at section 19(2)(d) to incorporate such a test. 57 A simple solution would be to replace the wording of section 19(2) (d) (a proportionate means of achieving a legitimate aim) with the words 53 See, for instance, Case C-13/05 Chacon Navas v Eurest Colectividades SA (2006). 54  'objectively justified by a legitimate aim and the means of achieving that aim are appropriate and necessary' , an amendment that would align the wording more closely with EU law. 58 Alternatively, the Supreme Court could take the so-far untried option of interpreting section 19(2)(d) as incorporating a strict test of necessity, rather than a test of what is (reasonably) necessary. We reluctantly admit, however, that this is extremely unlikely in view of the fact that the appeal courts have, on several occasions, rejected the imposition of a strict necessity test to justification. 59 The second option would be for the courts to further develop the Hampson test in a way which ensures that a robust application of proportionality is applied to the justification defence. Both of these options carry their own advantages and disadvantages. It may only be possible to remove discrimination by having a strict test of necessity since anything less gives the employer leeway to discriminate and blurs the cause of disparate impact. 60 From the evidence of the cases examined here, where the hypothetical application of a strict test of necessity resulted in a more favourable outcome for the claimant, this was due to the existence of suitable, less discriminatory alternatives. This suggests that if such a test were to be incorporated into the EqA, the protection afforded to claimants from indirect discrimination would be strengthened.
A further advantage of incorporating the Bilka test into UK law is its simplicity. The current test has led to considerable difficulties in interpretation since it leaves a great deal of discretion to tribunals. The advantage of a strict test of necessity is that, in reducing the discretion of an ET, any latent discrimination might be eliminated. Since employment tribunals are not generally composed of those groups who are most likely to experience discrimination, 61 there is a danger that judicial 'common sense' may reinforce, rather than challenge, discrimination. 62 This is a particular risk with indirect discrimination where the practices being considered may seem innocuous to those who are not personally troubled by them. A straightforward test of necessity is likely, therefore, to strengthen protection for claimants and to increase the consistency of decision-making at the level of the ET.
There are disadvantages, however, in incorporating the Bilka test into UK law. A strict test of business necessity is inflexible, and can be overly deferential to the needs and discretion of business. Also, while the Bilka formulation does not preclude any consideration of discriminatory impact, neither does it specifically encourage it and there is a risk that weighty discriminatory impacts will not be given sufficient consideration if this test is adopted (nevertheless, this is not inevitable since the Bilka test in fact has four separate elements to it, each of which needs to be considered: real need, appropriateness, necessity and proportionality, 63 although in reality the last part of the test is rarely highlighted). 64 Another potential difficulty is that under Bilka the question of justification is heavily dependent on the choice of legitimate aim. This leads to a risk in that an employer could tailor a legitimate aim to fit the Bilka requirements. This is a particular danger since courts and tribunals are increasingly willing to accept cost-saving as a legitimate aim. Although this was initially rejected by the courts (Hill and Stapleton v The Revenue Commissioners and Department of Finance), 65 in Cross v British Airways, it was accepted that, while cost-saving alone cannot constitute a legitimate aim, costs-saving plus further factors could do so. 66 This gives an employer even greater leeway in establishing a legitimate aim particularly in the current climate of austerity. This highlights a serious shortcoming of the Bilka test which is ultimately more deferential to the needs of business than to the aim of achieving substantive equality.
The lack of flexibility in the Bilka test can also be a problem, clearly demonstrated in Redfearn v Serco where the existence of alternative means prevented the dismissal of a BNP councillor from being justified under a strict test of necessity. This case highlights the need for courts to have the freedom to consider the societal implications of a measure.
A further consideration in relation to adopting the Bilka test would be the implications of the UK leaving the EU. If the UK courts failed to fully apply the jurisprudence of the CJEU prior to 'Brexit', it is unlikely that they will succumb to it now. The alternative option, which is for the UK courts to develop a more robust approach to existing anti-discrimination law, becomes even more compelling if the UK leaves the EU. Furthermore, as the UK courts have already considered discriminatory impact in a wider sense in cases such as JFS and Elias, this would be a natural development of UK case law. Encouraging this would help domestic anti-discrimination legislation to develop in a progressive way if the UK courts cease to be bound by the rulings of the CJEU. This would reduce the risk of UK anti-discrimination legislation being undermined by the UK's exit from the EU. In addition, as pointed out by Baker, this approach would be more consistent with the concept of justification under the European Convention on Human Rights. 67 Since proportionality in this context is about striking a balance between the infringement of a right and a legitimate aim, 68 adopting this approach in respect of domestic anti-discrimination legislation would ensure that these provisions develop in tandem with human rights legislation.

CONCLUSION
Our examination of EAT decisions shows that, where a robust application of proportionality is adopted, this has almost equivalent benefits for claimants as a strict test of necessity. In addition, as demonstrated in Redfearn, it gives the Court greater flexibility in cases where a strict test of necessity might produce a result which is out of kilter with the intentions of anti-discrimination legislation. A robust approach to proportionality requires the court to consider information regarding discriminatory impact, allowing the court to make an informed decision as to discriminatory effects. Furthermore, as this approach is evidence based, the decision-making process would have greater transparency, even in areas where the court has discretion.
The difficulty with this approach is that evidence appears to point to courts adopting varying degrees of intensity in the application of proportionality. 69 In addition, UK courts have not applied proportionality in the context of human rights any more robustly or consistently than they have in respect of anti-discrimination legislation. There is still considerable debate as to the appropriate 67 Baker n.11. 68 Belgian Linguistics Case (No 2) (1968) 1 EHRR 252. 69 See comments of the Lords in Hunag v SSHD [2007] UKHL 11. intensity of review within a proportionality assessment. Lord Hoffman, for instance, expressed the view that there is little difference between proportionality and Wednesbury 70 reasonableness. 71 Others, such as Rivers, state that the intensity of a proportionality review should be in parallel with the seriousness of the right infringed, 72 while Chan states that proportionality requires, at the very least, a baseline review for all rights. 73 Our evidence suggests that the EAT and ET do not consistently apply proportionality assessments with a high level of intensity. Our view is that in the context of justification the application of proportionality should include a consideration of the societal impact of discrimination. This would increase the intensity of review and ensure that decisions comply with the intentions of anti-discrimination legislation.
Our evidence suggests that the UK's current approach to justification results, for the most part, in a 'watered down' application of proportionality as necessity is qualified by reasonableness and the discriminatory impact is not adequately weighed. This is frustrating, as a robust application of proportionality could have a substantial positive impact on the effectiveness of anti-discrimination legislation. Furthermore, this approach could have some advantages over the application of a strict necessity test. It could, for example, provide a better balance between competing rights and discriminatory impacts; it would allow the courts to vary the intensity of review according to the weight of discriminatory impact; and the courts could more easily apply proportionality in accordance with the intentions of anti-discrimination provisions, providing there is sufficient evidence as to the overall effect of a discriminatory measure. Perhaps most importantly, it has greater potential to challenge discrimination than the Bilka test, since a significant discriminatory impact could outweigh business necessity.
Additionally, although the application of a robust approach to proportionality is complex and involves the court in determining questions of a socio-legalistic nature, this should present no more difficulties for tribunals than assessing the real and objective needs of an employer. Furthermore, although a full consideration of discriminatory impact would require the court or tribunal to consider extensive evidence of discriminatory impact, this information is readily available. High-quality, objective research on the effects of discrimination is available from organisations such as the Equality and Human Rights Commission and elsewhere. 74 Such evidence has recently been used to good effect in discrimination cases concerning welfare legislation. Cases such as R (SE v SSWP) 75 which concerned sexual discrimination in respect of a benefit cap and Hurley & Orrs v SSWP [2015] 76 demonstrate how a Court can make a more informed decision as to the legality of a discriminatory measure when its full impact is considered.
Hurley concerned a measure to restrict benefits for carers. This was deemed to be unjustified by the High Court on the grounds that the discrimination to carers was disproportionate. In this case, extensive evidence was provided to the court by organisations such as Carers UK. 77 This included information as to how much money carers saved the public purse-estimated to be about £119 billion pounds per annum. This evidence was crucial in outweighing the Government's legitimate aim which was to save money by encouraging carers to take up paid employment.
We suggest that representatives and claimants in cases concerning indirect discrimination routinely present wide-ranging evidence of discriminatory impact alongside helpful case law precedents such as Elias and JFS. This approach may act to encourage tribunals to adopt a more robust approach to proportionality and assist the development of the law in this area. Academics can also assist practitioners by producing high-quality research on the societal impact of discrimination. We suggest that this is necessary, as although this study has demonstrated that the UK is not fully compliant with EU standards with regard to justification, there is no indication that either the UK Government or the courts have any intention of amending the law. Therefore, although regrettable, it is unlikely that either of the two options identified in this study will be pursued unless concerned parties take action to initiate change.