Gibson & Ors v Sheffield City Council

This is an appeal against a decision of the Employment Appeal Tribunal (“EAT”), His Honour Judge Peter Clark presiding, dated 17 February 2009. The EAT dismissed an appeal by Mrs Lynn Gibson and others (“the appellants”) against a decision dated 26 March 2008 of an Employment Tribunal (“the Tribunal”) held at Sheffield, dismissing their claim against Sheffield City Council (“the Council”) under Section 1 of the Equal Pay Act 1970 (“the 1970 Act”). The Tribunal upheld the claim of other female employees against the Council and the Council’s appeal against that decision was dismissed by the EAT.

The appellants are employed by the Council as carers, a term which includes escorts, those assisting special needs children from home to school, care workers and school supervisors (formerly “dinner ladies”). The successful employees were cleaners. The Council have throughout accepted that the female claimants in these groups performed work rated as equivalent to that of their male comparators for the purposes of Section 1(2)(b) of the 1970 Act, a rating based on job ratings contained in the 1987 “White Book” in manual grades 1 to 5. The comparators were male street cleaning workers and male gardeners whose basic pay was 33.3% and 38% respectively higher than that of the women.

The issue was whether the Council had proved the absence of sex discrimination under section 1(3) of the 1970 Act. That issue has been determined in the Council’s favour without the Council being obliged to justify objectively the pay disparity. The Tribunal concluded that the pay differential was not ‘tainted by sex’, the expression used in Armstrong v Newcastle upon Tyne Hospital NHS Trust [2006] IRLR 124, to which reference will be made.

The pay differential has its origins in a bonus scheme for manual workers devised by the City Council in the 1960s. Following Government guidance that pay rises be related to increases in productivity, the Council employed management consultants to consider the operation of the City Engineers Department and report on “the different types of work and the types of bonus that could be introduced as well as the potential savings to be made by increases in productivity and the consequential reorganisation or reduction in staff” (Tribunal decision, paragraph 3.13). The consultants recommended (paragraph 3.17) that savings could be achieved “by providing productivity payments and thereby reducing labour, machinery and associated costs”. The Tribunal accepted (paragraph 3.23) that the introduction of bonus schemes for the comparator groups was “a true attempt to relate wages to productivity”. “Genuine schemes were introduced by which productivity at agreed levels produced entitlement to an agreed level of productivity bonus” (paragraph 3.35). “The aim was to improve productivity, recruitment, retention and increase pay without additional costs” (paragraph 3.37). “Productivity was measured against work study models and if 100 performance was achieved 33.3% bonus paid, for 110 performance 38%. Productivity was monitored by foreman and supervisors, bonus and timesheets completed and checked by pay clerks” (paragraph 3.38).

In the 1980s, the Council conducted a procedure described as “stabilisation” of the bonus scheme. The Tribunal referred (paragraph 3.42) to an agreement in relation to the mechanical sweeper operators made in February 1987. “The terms of this include a movement to flexible working including Saturdays and Sundays and a 39 hour week. All overtime was to be eliminated and the aim was to avoid fluctuation in wages by shift scheduling and an agreed target. There is an expectation of a high degree of flexibility in working practices and as a result the employees will have the benefit of stable and predictable wages”. The agreements reflect “a continuing element of an operational efficiency payment in return for 100 productivity, the payment being 33.3%”. The Tribunal concluded that “on balance . . . these agreements reflected agreement as to achieved productivity as an average for which, in addition to other agreements as to flexibility etc., the respondent agreed to maintain an average bonus of 33.3%, i.e. 100 performance” (paragraph 3.49).

The Tribunal noted (paragraph 3.53) that “there is no provision for reduction of operational efficiency payments in the event of a group failure to meet targets and therefore the only sanctions available are monitoring of the performance together with disciplinary action for failure to comply with required standards” (3.53). “It was part and parcel of the stabilisation and the balancing of the payments means that there is effectively no reduction or increase in pay . . . After some 10 years or so of the bonus schemes productivity levels had moved up to levels reflecting standard (100) performance or better and therefore this was reflected by giving an average of 100 performance for the street cleaning employees and 110 performance for grounds maintenance based on average productivity” (paragraph 3.55). “We were told that the bonus in relation to the gardening workers at 38% was based on average productivity levels achieved over a previous period of 2-3 years. We find this to be the case”. “We find that 110 performance had been identified as being achieved and therefore they received 38% bonus” (paragraphs 3.57-3.58).

The Tribunal heard evidence from some of the male comparators. It found that written terms and conditions of employment did not identify the productivity element of the wage as a separate figure and that there was no identification of any bonus element of pay in the payslips (paragraph 3.71 and 3.73).

No attempt was made to identify for new starters that they had pay entitlement based on a bonus. There were set targets for them to achieve but no expectation to show increases in productivity (paragraph 3.75). None of the comparators believed that their pay would be affected by a failure to meet targets (paragraph 3.78). The Tribunal found that very little qualitative assessment was required to determine whether work had been performed at levels required to meet the level of pay given (paragraph 3.79). The bonus had been introduced to increase productivity and thereafter preserve it. The level of productivity was maintained and the bonus preserved (paragraph 3.80).

The Tribunal noted that a job evaluation was conducted in 1987/88. As appears from a National Joint Council (“NJC”) for Local Authority Services (Manual Workers) report the assessment which led to the work of the appellants being assessed as of equal value to those of employees receiving these substantial bonuses was detailed and systematic. The Tribunal found that there was no indication that productivity was factored into job evaluations (paragraph 3.83). Having referred to a 1998 report of the Bonus Technical Working Group of NJC, the Tribunal stated, at paragraph 3.87:

“The key objective of single status is set out as to provide a fast and non-discriminatory pay and grading structure at local level for all employees. There is an expectation on Councils and trade unions to check that bonus arrangements remain appropriate, non-discriminatory and effective. At page 676 is stated, ‘A key element in such an exercise will be to ensure that each bonus pay scheme is applied in a manner that complies with equal pay and sex discrimination legislation’. There are statistics at page 677 indicating the majority of the manual workforce is female (515,202 of 651,669), 57.3% of full-time men received bonus compared to 6.9% of full-time women and road worker groups followed by gardeners and drivers were the top 3 in a ‘bonus league’ and all are said to be male dominated groups.”

The Tribunal noted, at 3.90, a statement in the report:

“There is a further statement that if female groups are not given access to productivity bonus because they are already fully productive the employer is unlikely to be able to justify male groups receiving bonus irrespective of productivity levels or not giving female workers the same bonus for doing work of equal value at equivalent levels of productivity. This is effectively a warning from the National negotiating body as to the equal pay challenge inherent in the single status agreement.”

The employers were thus on notice of the potentially discriminatory effect of bonus schemes and that the schemes are the subject of “continuing negotiations between the management and trade union side” (paragraph 3.103).

Under the heading “Statistics”, the Tribunal stated that “it is clear that of the workers engaged by the [council] the vast majority in receipt of bonus payments are male. This is true no matter what analysis is made of whichever pool and whichever of the 3 dates of statistics are chosen”:

“3.114

From these can be seen that the vast majority of employees in the claimant group jobs are female. The figures vary from 84.7% to 100% but 10 out of the 14 are above 95%. Of the comparator jobs 4 have no female employees, the other 3 are 7.7%, 11.1% and 12.5%. The statistics so far as 1996 is concerned show a majority of full-time workers as male (92.5%). By 1996 56,623 of 146,626 full-time staff are female. 79.1% of staff on manual worker grades are female. 57.3% of full-time males are on bonus and only 6.9% of females, the equivalent of 39.4% of all male employees on bonus and only 3.1% of females. In 1998 75.4% of male manual and craft workers are on bonus and only 1.6% of females. The information is analysed by directorate. By 2006 the comparison remains as the comparator groups being mainly at 100% male (only 3 differ).

3.115

The impact of these statistics is that it is clear that the bonus has been given to predominantly male workers, that this has continued to 2006 and was present when bonuses were first implemented.”

These findings are supported by a detailed schedule of agreed statistics.

The Tribunal considered stereotypical assumptions:

“3.116

We consider between us that it is correct that there were stereotypical assumptions that the Claimants’ work was essentially ‘women’s work’ and the comparators work considered ‘mans work’. We consider this in relation to 1988 first of all and consider the same to be true in the 1960’s and 1970’s that cleaning and looking after people was stereotypically viewed as women’s work. By contrast using a barrow or machine to clear the streets or to drive a vehicle cleaning roads, using machinery to cut grass or hedges or doing general garden tasks was probably perceived as mans work. We considered the position in 1999 and feel that the same was true. We feel also that this is reflected in the Respondent’s evidence that there was difficulty in encouraging for example men into caring roles or women to the comparator roles. This is not to say that we ignore the fact that there may be women and men engaged on each of the roles nor do we believe the respondent wanted only to recruit men or women to the roles but that as a general proposition socially the jobs of the Claimants were viewed as ‘women’s work’ and the comparators’ as ‘men’s work’.”

Against that background, the Tribunal considered in great detail whether a bonus scheme based on productivity could be introduced for the appellants. They did so, as is common ground, on the basis stated at paragraph 3.136:

“The ‘care workers’ impressed as a dedicated group of individuals who, as with the supervisors and escorts, enjoyed the nature of their work and put the interests of the client or children first.”

The Tribunal’s answers to the questions posed were resoundingly in the negative:

“It would not be appropriate . . . when the essence of caring is that the clients needs are met by the carer and therefore the pace of work is beyond control” (paragraph 3.121).

“This in our view underlines the nature of providing personal care as not being an objectively measurable task by input such as can be the basis of a calculable bonus reward” (paragraph 3.122).

“We find again that the tasks were very much dependent on the level of co-operation of the children. We do not believe it was at any time feasible to provide an incentive to work more quickly or efficiently. It was not a series of repetitive tasks which could be measured or counted” (School supervisors, paragraph 3.141).

“Our finding is that the tasks for the carers in terms of outputs are not measurable and it would not be practicable for such a bonus scheme to be applied” (Escorts, paragraph 3.144).

“So far as the supervisors and escorts were concerned we did not feel that there was any prospect of introducing bonuses in relation to reducing the time taken to perform tasks by more efficient working” (paragraph 3.146).

Having reaffirmed that the appellants’ work had been assessed as being of equal value to that of their comparators, the Tribunal also confirmed their view that the bonus scheme was a genuine scheme at its inception and “the conversion of the variable bonus recognised and reflected achieved production levels and provided a stable reward for continuing to work at that level” (paragraph 5.3). From the Tribunal’s detailed consideration of the evidence it emerges that:

The Tribunal expressed a general conclusion about the possibility of a bonus scheme for the carers:

“5.4

We have found that such a scheme could not be applied to school supervisors or to school escorts. We have also found that such a scheme could not apply to persons providing personal care.

5.5

We have no evidence of a positive decision being made not to apply a productivity bonus to any of these grades. However, the culture within the care sphere was that such payments would not be conducive to personal care. There is independent evidence of this. We have set out our reasons as to why we believe such a scheme would not be possible which includes; the job is not a series of mechanical repetitive tasks, it cannot be objectively assessed, there is no regular work input and it is determined by the recipient, the recipient pays a contribution to care costs, the workers themselves could not anticipate how such a scheme could work and there would be a risk of workers ‘re-writing’ care plans and a perceived risks of attribution of errors to rushing caused by the need to earn a bonus, there had never been a known example of bonuses in such work areas.”

The Tribunal also concluded, by way of contrast, and for reasons set out in considerable detail, that a productivity bonus scheme could apply to the cleaners. At paragraph 5.19, the Tribunal concluded, in relation to cleaners:

“There was indirect discrimination. A pay policy was applied which adversely affects women disproportionately when compared with their male counterparts. We do not believe the respondent has discounted any possible relationship between gender and the provision of bonuses.”

The Tribunal went on to find that the payment of bonus to comparators and not to the cleaners was not objectively justified. Cost was the real reason for the failure to achieve equality (paragraph 5.24).

Having considered, in considerable detail, the statistics, the perceptions, the work done, the development of the bonus scheme and its possible applicability to the appellants, general conclusions in relation to the appellants were expressed concisely by the Tribunal:

“5.7

We consider therefore that the reason the care workers, meal supervisors (and senior supervisor) and escorts do not receive a bonus is made out. The respondent has therefore proved in our view the reason why it pays a bonus to the comparators but not to the claimants in those roles . . .

5.8

The reason is the need to provide payments for increased productivity which it cannot and does not need to make to these claimants.

5.9

This is a reason which is not the gender of the recipients of the bonus or the claimants. In our view the respondent has proved that the reason is not the reason of the sex of the group of workers. The respondent having proved this is not required to objectively justify the payments. They have proved there is no discrimination.

. . .

5.26

By reference to the legislation we believe the reason why the male comparators receive a bonus is in return for productivity in relation to outcomes which are measurable. This productivity is achieved by more flexible efficient working which has been kept up to date by the various pressures on the part of the respondent employing the comparators. It is a genuine scheme. It is material in that it is there to provide efficiency of production. It is unrelated to the gender of the recipients.

5.27

. . . [The Council] has shown it could not apply such a bonus to the claimants who are carers, escorts or lunchtime supervisors . . .”

In upholding the Tribunal, the EAT’s reasoning was consistent with that of the Tribunal:

“34.

We return to the principal issue. It seems to us that, applying Armstrong as analysed by the President in Surtees without disapproval by the Court of Appeal in that case, even assuming that disparate impact was made out, it remained open to the Employment Tribunal to find that the difference in carers’ pay was not sex tainted and they so found, permissibly, on the basis that (a) the consolidated bonus in the male comparators’ pay was genuinely referable to productivity and (b) that a similar productivity bonus was inappropriate for the carers. In this way the Council negatived the taint of sex in the pay differential and thus were not required to objectively justify that differential.

35.

Having rejected Ms Gill’s principal submission in this appeal we can deal shortly with two further grounds. First, she relies upon the statistical impact on the carers and their comparators, together with the Employment Tribunal’s finding of ‘stereotypical assumptions’ (para. 3.116) for the proposition that the Employment Tribunal failed to look behind the reason found for the difference in pay to its impact on the disadvantaged group. With respect, that seems to us to be another way of seeking to circumvent the effect of Armstrong. Even where there is disparate impact which has an adverse effect on the women’s group it remains open to an employer to prove that the pay differential is not sex tainted. On the facts, that is what the Council succeeded in doing in relation to the carers (but not the cleaners).”

The Tribunal did not mention Armstrong in their reasoning but appear to have relied on it, as has the EAT.

The further ground was an allegedly perverse finding of fact (paragraph 3.148) in relation to the bonus scheme and need not be pursued. I will defer comment upon the Tribunal and EAT judgments until after I have considered the authorities mentioned and other authorities.

Requirements for equal pay for male and female workers have their origins in Article 141 of the Treaty of Rome:

“1.

Each Member State shall ensure that the principle of equal pay for male and female workers for equal work or work of equal value is applied.”

In Enderby v Frenchay Health Authority [1994] ICR 112 a predominantly female occupational group, speech therapists, compared their pay with two predominantly male occupational groups, pharmacists and clinical psychologists. The reason for the difference in pay was that the pay of the different occupational groups was determined under distinct collective bargaining arrangements. The employers relied on the defence in section 1(3) of the 1970 Act:

“An equality clause . . . shall not operate in relation to a variation between the woman’s contract and man’s contract if the employer proves that the variation is genuinely due to a material factor which is not the difference of sex . . .”

In the present case the Tribunal and the EAT held that the reason for the disparity in pay was not the difference of sex so that the Council was not required to justify the disparity objectively.

The ECJ answered questions posed in Enderby:

“(1)

Where significant statistics disclose an appreciable difference in pay between two jobs of equal value, one of which is carried out almost exclusively by women and the other predominantly by men, Article [141] of the EEC Treaty requires the employer to show that that difference is based on objectively justified factors unrelated to any discrimination on grounds of sex.

(2)

The fact that the respective rates of pay of two jobs of equal value, one carried out almost exclusively by women and the other predominantly by men, were arrived at by collective bargaining processes which, although carried out by the same parties, are distinct, and, taken separately, have in themselves no discriminatory effect, is not sufficient objective justification for the difference in pay between those two jobs.”

Directive 97/80/EC attempted to incorporate ECJ case law:

“1.

For the purposes of this Directive, the principle of equal treatment shall mean that there shall be no discrimination whatsoever based on sex, either directly or indirectly.

2.

For the purposes of the principle of equal treatment referred to in paragraph 1, indirect discrimination shall exist where an apparently neutral provision, criterion or practice disadvantages a substantially higher proportion of the members of one sex unless that provision, criterion or practice is appropriate and necessary and can be justified by objective factors unrelated to sex.”

It is not necessary for present purposes, to consider a consolidated Directive 206/54/EC, which was not in force at the material time.

The effect of the 1970 Act and the Directives was considered by Lord Nicholls of Birkenhead in Glasgow City Council v Marshall [2000] ICR 196. Lord Nicholls, at page 198D, began his speech, with which the other members of the Committee agreed, by stating what was in issue. Section 1(3) of the 1970 Act provides a “material factor” defence to an equal pay claim:

“This defence avails an employer if he proves that the difference in pay is genuinely due to a material factor which is not the difference of sex and that factor is a material difference between the woman’s case and the man’s case. This appeal raises, once more, the much-discussed question of the proper interpretation of this subsection.”

Having recited the facts Lord Nicholls stated:

“Given that instructors are engaged on like work with their teacher comparators, and given also that instructors are paid far less than the teachers, the onus was on the education authorities to establish a defence under section 1(3).”

Lord Nicholls expressed his conclusions on that issue but did so in a case in which it was common ground, including by reference to statistics, that there had been no sex discrimination (page 200F). Both groups were over 95% female. The Tribunal’s finding in favour of the applicants was rejected. The finding was that “even in a case where the absence of sex discrimination was demonstrated, some good and sufficient reason must exist for the variation in pay. If none was proved, the claim succeeded” (page 201B).

Lord Nicholls stated, at page 202 C – E:

“This [the Tribunal’s] approach would mean that in a case where there is no suggestion of sex discrimination, the equality clause would still operate. That would be difficult to reconcile with the gender-related elements of the statutory equality clause. The equality clause is concerned with variations in pay or conditions between a woman doing like work with a man and vice versa. But if the equality clause were to operate where no sex discrimination is involved, the statutory starting point of a gender-based comparison would become largely meaningless. On this interpretation of the Act, what matters is not sex discrimination. What matters is whether, within one establishment, there is a variation in pay or conditions between one employee doing like work with another employee. The sex of the employees would be neither here nor there, save that to get the claim off the ground the chosen comparator must be of the opposite sex. On this interpretation the Act could be called into operation whenever mixed groups of workers are paid differently but are engaged on work of equal value. In such a case the statutory equality clause would operate even when the pay differences are demonstratively free from any taint of sex discrimination.”

It was in the absence of sex discrimination in the particular case that, at page 202F, Lord Nicholls stated his conclusions on the general issue raised:

“The scheme of the Act is that a rebuttable presumption of sex discrimination arises once the gender based comparison shows that a woman, doing like work or work rated as equivalent or work of equal value to that of a man, is being paid or treated less favourably than the man. The variation between her contract and the man's contract is presumed to be due to the difference of sex. The burden passes to the employer to show that the explanation for the variation is not tainted with sex. In order to discharge this burden the employer must satisfy the tribunal on several matters. First, that the proffered explanation, or reason, is genuine, and not a sham or pretence. Second, that the less favourable treatment is due to this reason. The factor relied upon must be the cause of the disparity. In this regard, and in this sense, the factor must be a 'material' factor, that is, a significant and relevant factor. Third, that the reason is not 'the difference of sex'. This phrase is apt to embrace any form of sex discrimination, whether direct or indirect . . .

When section 1 is thus analysed, it is apparent that an employer who satisfies the third of these requirements is under no obligation to prove a 'good' reason for the pay disparity. In order to fulfil the third requirement he must prove the absence of sex discrimination, direct or indirect. If there is any evidence of sex discrimination, such as evidence that the difference in pay has a disparately adverse impact on women, the employer will be called upon to satisfy the Tribunal that the difference in pay is objectively justifiable. But if the employer proves the absence of sex discrimination he is not obliged to justify the pay disparity.”

Before the House, counsel for the applicants had sought to introduce fresh statistics demonstrating that, in the region as a whole, a significant percentage of the category receiving the higher wage (teachers) were men (page 205 A – B). Lord Nicholls stated, at page 205C:

“Whether a pay disparity has a disparately adverse impact on women is primarily a question of fact. . . . This issue has never previously been suggested. In all conscience, these proceedings have long passed the stage at which the applicants should be permitted to introduce fresh issues of fact.”

In Wilson v Health & Safety Executive Quality & Human Rights Commission (Intervener) [2009] EWCA 1074, Arden LJ stated, at paragraph 35:

“The jurisprudence of the Court of Justice shows that equal pay for equal work or work of equal value is a fundamental principle of the Treaty. It is one of the values of the European Union and its achievement is important to the social and economic goals of the European Union.”

Arden LJ added, at paragraph 64, that:

“The right to equal pay ‘expresses an ideal of social progress and human dignity’.”

In the present case, there is no doubt:

(a)

The work of the appellants and that of the male comparators was work of equal value under section 1(2)(b) and section 1(5) of the 1970 Act,

(b)

The male comparators were paid substantially more than the appellants,

(c)

The appellants are women and the pay difference, as the statistics cited by the Tribunal demonstrate, disadvantaged a substantially higher proportion of women.

The reasoning of the Tribunal is attractively simple. To apply a productivity bonus to the carers would not be conducive to personal care. It could not be applied to carers. A productivity bonus cannot be paid to carers. Therefore, it does not need to be paid to them. That reason is unrelated to the gender of the recipients. The employers have proved there is no discrimination (paragraphs 5.7 to 5.9 and 5.26). No obligation to justify the disparity objectively arose. The Tribunals appear to have relied on the decision of this court in Armstrong, considered in subsequent cases.

Armstrong is not an easy case. The claimants were female domestic ancillary workers and sought to compare themselves with male ancillary workers who were subject to bonus arrangements. The Tribunal (upheld by the EAT) found that no defence under Section 1(3) of the 1960 Act was available. Justifications for that decision included factors found to be tainted by sex. This court reversed the decision and, by a majority as to remission, remitted the issue of disparate adverse impact to the Tribunal. Other issues were involved but potentially relevant to the present appeal are statements made by Arden LJ and Buxton LJ (with whom Latham LJ agreed) which purport to give general guidance on the operation of Section 1(3) of the 1970 Act. Arden LJ stated:

“32.

I have set out in paragraph 17 above the well-known passage from the speech of Lord Nicholls in the Marshall case. (The remainder of the House agreed with his speech.) That passage sets out a step by step guide to proving a genuine material factor defence. For the purposes of this appeal, the steps can be summarised as follows:

(1)

the complainant must produce a gender-based comparison showing that women doing like work, or work rated as equivalent or work of equal value to that of men, are being paid or treated less favourably than men. If the complainant can produce a gender-based comparison of this kind, a rebuttable presumption of sex discrimination arises.

(2)

the employer must then show that the variation between the woman's contract and the man's contract is not tainted with sex, that is, that it is genuinely due to a material factor which is not the difference of sex. To do this, the employer must show each of the following matters:

(a)

that the explanation for the variation is genuine,

(b)

that the more favourable treatment of the man is due to that reason, and

(c)

that the reason is not the difference of sex.

(3)

if, but only if, the employer cannot show that the reason was not due to the difference of sex, he must show objective justification for the disparity between the woman's contract and the man's contract.

33.

It follows from the Marshall case that there is no need for an employer to provide justification for a disparity unless the disparity is due to sex discrimination. Miss Tether does not submit any different principle applies by virtue of Article 141.

34.

In the Marshall case, Lord Nicholls uses the words "disparately adverse effect". He held that evidence that a difference in pay had such an effect on women could be evidence of sex discrimination. He did not, however, hold that the mere fact that there was a disparately adverse effect was itself sex discrimination. In the Marshall case, Lord Nicholls used the phrase "disparately adverse effect" to denote the trigger at which the rebuttable presumption of sex discrimination would arise under step 1 mentioned above.”

Buxton LJ stated, at paragraph 103:

“The figures, if they demonstrate a disproportionate or (what seems to be the same thing) disparately adverse effect on female employees, thus open the door to a claim of indirect discrimination, by placing on the employer the burden of proving that the variation in terms between the woman's and the man's contract is "genuinely due to a material factor which is not the difference of sex": Equal Pay Act 1970, section 1(3). How he discharges that burden was the major matter of dispute in this case.”

At paragraph 110, Buxton LJ, apparently relying on the last sentence of the passage cited from Marshall, approved the approach of Arden LJ:

“As Lord Nicholls said at the end of the passage from Glasgow City Council v Marshall . . . if the employer proves the absence of sex discrimination he is not obliged to justify the pay disparity. That is the basis of the step by step approach explained by Arden LJ in her paragraph 32 above. Once disparate adverse impact has been established, the burden passes to the employer in respect of two issues. First, that the difference between the man's and the woman's contract is not discriminatory, in the sense of being attributable to a difference of gender. Second, if the employer cannot show that the difference in treatment was not attributable to a difference of gender he must then demonstrate that there was nonetheless an objective justification for the difference between the woman's and the man's contract. I deal with those issues in turn.”

Thus, the Tribunals in the present case concluded, the difference between the appellants’ contract and the comparators’ contract is genuinely due to a material factor which is not the difference of sex because the nature of the appellants’ work is such that bonus schemes cannot be applied to it.

In Villalba v Merrill Lynch & Co Inc & Ors [2007] ICR 469 the EAT, Elias J (President) presiding, referred to the EAT’s difficulty with Armstrong but made a clear statement at paragraph 113:

“In effect, therefore, Enderby establishes that statistics alone may in a sufficiently powerful case create an irrebuttable presumption of prima facie indirect sex discrimination. Once there is statistically relevant and material evidence to demonstrate that a group is in fact being adversely affected on sex grounds, that will oblige the employer to justify the pay arrangements. It is not enough in those circumstances for the employer to demonstrate that the arrangements have resulted without any direct sex discrimination being practised of any kind. In effect the court is holding that there must somewhere have been some element of indirect discrimination - in other words there is an assumed indirect discrimination. The statistics demonstrate a sufficiently marked adverse impact to constitute a prima facie case requiring not merely an explanation that the difference is not caused directly by sex, but in addition an objective justification. The statistics must at least show that it is reasonable to infer that the treatment of the disadvantaged group must have resulted from some factor or combination of factors which impinge adversely on women because of their sex, even though no obvious feature causing this disparate treatment can be identified, and indeed even though the employer has apparently demonstrated to the contrary.”

In Middlesbrough Borough Council v Surtees & Ors [2007] ICR 1644, however, the EAT, Elias J (President) presiding, stated that the EAT was bound by the decision in Armstrong and accepted, at paragraph 48, that “if the employer adduces evidence which satisfies the Tribunal that the variation in pay is not for sex reasons, that objective has been satisfied and there is no reason in policy or logic for requiring a non-sex based distinction to be objectively justified”. In a lengthy analysis at paragraphs 49 to 55, Elias J postulated circumstances in which the Armstrong principle might “exceptionally” apply.

Giving the judgment of the court in Redcar & Cleveland Borough Council v Bainbridge & Ors (Equality & Human Rights Commission Intervening); Middlesbrough Borough Council v Surtees & Ors [2009] ICR 133, Mummery LJ described the reasoning of the EAT in Surtees:

“In its Middlesbrough decision [2007] ICR 1644 the Employment Appeal Tribunal considered Armstrong in some detail and concluded that it was correctly decided, but that it would be of limited effect in practice. The Appeal Tribunal considered that, if a tribunal found that there had been disparate adverse impact (at least if the statistical evidence was convincing), it would usually be impossible in practical terms for the employer to show that the pay disparity was not related to the difference in sex. In particular, in an Enderby-type case, where the disadvantaged group comprised all or almost all women and the advantaged group were all or almost all men, it would be impossible in practical terms for the employer to show that the pay differential was not related to the difference of sex. Thus, although, in theory, it was right to say that an employer could still show that the pay differential was not due to the difference of sex, even in the face of evidence of disparate adverse impact, it would only be possible in a case where the statistical evidence was not very strong or convincing.”

The court did not decide the issue but in the following paragraph, Mummery LJ stated that the court was “inclined to think that that analysis of the decision is correct”.

The EAT, Elias J (President) presiding, again considered Armstrong in Cumbria County Council v Dow & Ors [2008] IRLR 91. At paragraph 108, the EAT accepted the proposition that “it is only if the employer fails to satisfy the Tribunal that the discrimination is not sex tainted that the question of justification arises”. At paragraph 119, the EAT held that “the Tribunal was fully entitled to find that the employers have failed to discharge that burden here”.

At paragraph 114, the reasoning of Elias J appears to deprive the Armstrong principle of any application in circumstances such as the present:

“It does seem to us that Mr Jeans [for the employer] is in fact running two inconsistent arguments. He has submitted that it was not possible to confer similar productivity benefits on any of the claimant groups and indeed, appeals against the Tribunal's finding that it was possible to adopt productivity schemes for caterers and cleaners. If he is right about that, then in our view the only conceivable finding is that there was sex tainting. If the benefit is given only to those who perform traditionally male jobs and cannot be conferred on the claimants doing traditionally women's work, then prima facie indirect discrimination inevitably arises. On that premise, only the predominantly male groups can benefit from the way the pay arrangements are structured. It is akin to paying more to full timers than part timers. The differential may be justified, but the need to justify plainly arises.”

A similar statement was made by the EAT, Elias J (President) presiding, in Coventry City Council v Nicholls & Ors [2009] IRLR 345, at paragraph 46:

“Similarly here: if it is possible to make a payment only to an exclusively or almost exclusively male group, because of particular features of their job not shared by the female claimants, then it necessarily involves a form of prima facie indirect discrimination against those women. The payment is being made by reference to characteristics of a job which in practice are held by job holders who are predominantly of one sex only. That does not, of course, mean that to pay such bonuses is necessarily unlawful, but it will be unless it can be justified.”

We have been referred to Court of Appeal decisions pre-dating Armstrong in which the Armstrong sophistication did not feature. In Home Office v Bailey [2005] ICR 1057, Peter Gibson LJ, with whom Sir Martin Nourse agreed, stated, at paragraph 28:

“Provided that the Employment Tribunal is satisfied as to the validity of the statistics and the appropriateness of their use, in my judgment it is free to use that statistical approach in order to determine whether there has been prima facie discrimination. In particular I do not see anything in Enderby that precludes such approach.

29.

. . . In each case the Employment Tribunal is concerned to determine whether what on its face is a gender-neutral practice may be disguising the fact that female employees are being disadvantaged as compared with male employees to an extent that signifies that the disparity is prima facie attributable to a difference of sex.”

For the appellant, Mr Linden QC submitted that Armstrong was decided per incuriam. In any event, the relevant remarks were obiter, and should be ignored. They were inconsistent with decisions of this court such as Bailey and were contrary to Enderby. They were, in any event, of such limited potential application that they can be ignored in the present case where the statistics, which may create a prima facie case, were clear and compelling. I say now that I do not accept that the decision in Armstrong was reached per incuriam.

Mr Linden rightly accepts that there may be cases in which analysis of the statistics demonstrates that they are not for material purposes significant or that they are fortuitous or apply to the wrong groups. Where they are cogent, and as on the present facts compelling as showing a disparatively adverse impact on women, the employer is required to show an objective justification for the difference in pay.

Mr Linden also relied on the principle that the 1970 Act “has to be construed so far as possible to work harmoniously both with the Sex Discrimination Act 1975 and Article 119” (Strathclyde Regional Council v Wallis [1998] ICR 205, at 212, per Lord Browne-Wilkinson). The procedure under the 1970 Act should not be complicated by the Armstrong sophistication, it was submitted.

For the Council, Miss Lang QC relied on the reasoning of the Tribunal summarised at paragraphs 32 and 34 above. She submitted that Armstrong is binding. The Tribunal reached a conclusion it was entitled to reach on the basis of Armstrong, which is consistent with Enderby, Marshall and Court of Appeal decisions. Lord Nicholls in Marshall kept open the possibility that an employer could establish that there was no sex discrimination and, if the employer did so, the obligation to justify the disparity objectively does not arise. The Council did so in this case to the satisfaction of the Tribunal.

The material factor causing the differences in pay was not tainted with sex, it was submitted. The reason was succinctly stated, as it could be, and is sound. The productivity of the men is measurable, that of the women is not. Increased payments for productivity cannot and need not be made to the women.

The productivity bonus has created a disparity between the pay of men and women though that was not the intention when introducing it. It was intended to improve productivity in the men’s work, which was unsatisfactory. Following the payment of substantial bonuses, productivity levels by the men of 100 and 110 have been attained and maintained. However, there has been no suggestion at any time other than that the women are doing work of equal value and are achieving an equivalent level of efficiency. They have achieved efficiency without bonuses; the bonuses for men were necessary to make the men efficient. The difference in treatment, as frankly stated by the Tribunal, is that the men’s work can be measured to provide a benchmark for productivity and the women’s work cannot. The Tribunal found that it follows that the Council need not pay equal wages. There is no further reasoning.

That measurability is the key is underlined in the Tribunal’s paragraph 5.26, cited at paragraph 17 above. What follows in that paragraph is no more than a statement that the bonus scheme has achieved its object of making the comparators efficient. The carers have always been efficient, in the Tribunal’s view.

I find the approach in Armstrong very difficult to reconcile with the first of the answers by the ECJ in Enderby, cited at paragraph 22 above, the second paragraph of Directive 97/80/EC and with Marshall. Marshall was concerned with analysing the section 1(3) defence (page 198D and page 200E) but Lord Nicholls did so in a context in which sex discrimination was absent (200F, 201B and F). Lord Nicholls stated the three requirements the employer must then meet, elaborating on the third of them. Read in that context, I do not consider that the last sentence in the passage cited has the significance attributed to it by Buxton LJ in paragraph 110 of Armstrong (paragraph 34 above).

In Marshall, sex discrimination had not been alleged. The applicants had sought to rely on a line of reasoning advanced (notwithstanding the “gender-related elements of the statutory equality clause”) and decisively rejected by Lord Nicholls at page 202 C – E. In the present case, it is conclusively established that “the difference in pay has a disparately adverse impact on women”. There is evidence, on Lord Nicholls’s formulation, of sex discrimination. By contrast, in Marshall, there was, on the evidence, an absence of sex discrimination and the last sentence in Lord Nicholls’s analysis was therefore apt, which in the present case it is not. The House did not permit the introduction of fresh statistics which might have established “the disparately adverse impact on women” and thereby placed the burden of justification on the employers.

However, I am content to conclude only that Armstrong has no application on the present facts, given the clear and compelling statistics. I have considered the valiant attempts of Elias J, at paragraphs 51 to 55 in Surtees, to postulate situations in which the Armstrong approach might apply. I am not at present convinced that it has any practical application or that it adds anything to the analysis required of issues arising under section 1 of the 1970 Act. The approach of Elias J at paragraph 114 in Dow, and paragraph 46 in Nicholls, appears to me to be applicable in the present case and to be correct.

The effect of the productivity bonus, now stabilised, is discriminatory. A sexual taint is present. The impossibility of applying the productivity bonus to women’s work, carefully reasoned by the Tribunal, is genuine enough but that does not remove the sexual taint from the operation of the scheme. The scheme has a disparately adverse effect on women’s work as compared with men’s work and the sexual taint is present. It must be justified objectively if the employers are to succeed. The opportunity to justify is a sufficient protection for employers in circumstances such as the present.

I would allow this appeal. Mr Linden submitted that, in that event, there is no need to remit the case to the Tribunal because there is no real prospect that the Council will be able to justify the disparity objectively. In my judgment, it should be remitted to allow the employers the opportunity, if they seek to take it, to attempt such justification.

I agree with the judgment of Pill LJ. I agree that, on the facts of this case, as clearly set out in his judgment, it was perverse for the employment tribunal (ET) to hold that the pay differentials between the carers and the male comparators were not ‘tainted by sex’ and did not require objective justification. It follows that the matter must be remitted to the ET for consideration of the employer’s case on objective justification.

I add a few words of my own because I wish to say a little more about the case of Armstrong v Newcastle Upon Tyne NHS Hospital Trust [2006] IRLR 124.

This is the second appeal in which I have been part of a constitution required to consider a decision in which the meaning, scope and correctness of the decision in Armstrong has been argued. The first was Middlesbrough Borough Council v Surtees [2009] ICR 133 (Surtees) where, at paragraphs 57 to 60 of the judgment of the Court (Mummery, Smith LJJ and Lindsay J), we considered that it was not necessary to reach any definite conclusion about Armstrong. We were, however, of the view that it was certainly binding on us and was probably correctly decided, although of limited application.

A similar situation arises in the present appeal, where, although it is plain that the ET and EAT have considered Armstrong, Pill LJ has not found it necessary to reach any definite conclusions about it save to say that it was not decided per incuriam and is therefore binding on this Court. I agree with him about that. So, unless and until the Supreme Court says otherwise, it is open to an employer to avoid the need for objective justification if he can show that, notwithstanding that statistics have been produced which show that the pay practice in question has an adverse impact on women, that pay practice is not sex-tainted. I wish to say more because, on further consideration of the decision, I am convinced that it is correct. However, it is apparent from this appeal (and others) that the way in which an employer can show that the pay practice is not sex-tainted is readily misunderstood.

I think that the problem is that tribunals fall into the trap, as this tribunal did, of thinking that, provided there is an explanation for the difference of pay which is not directly discriminatory, that will suffice to show that the pay practice in question is not sex-tainted.

I wish first to explain why I have concluded that Armstrong was correctly decided. I start by drawing attention to the difference of wording in section 1(3) of the Equal Pay Act 1970 (EqPA) and the indirect discrimination provisions of the Sex Discrimination Act (SDA) 1975. Section 1(3) of the EqPA provides a defence to the employer who shows that the pay differential between the man and the woman is ‘genuinely due to a material factor which is not the difference of sex’. There is an express opportunity for the employer to defend himself by showing that the pay differential is not due to the difference of sex. The EqPA was drafted at a time when the concept of indirect discrimination had not been imported into the law of England and Wales. It was clearly drafted with cases of alleged direct discrimination in mind. Yet, as is clear from Lord Nicholls’ speech in Marshall (see the citation in paragraph 24 above), the genuine material factor defence will not be satisfied if the pay practice is affected by either direct or indirect sex discrimination. As Lord Nicholls says, the employer must prove the absence of direct or indirect discrimination.

Where direct discrimination is in issue, the employer will have to show that the pay practice in question does not discriminate against women qua women. If it does, it is unlawful and cannot be justified.

Lord Nicholls also says that the employer will have to show that the pay practice does not discriminate indirectly against women (note that Nelson v Carillion Services says that the claimant must prove that) and that, if he fails to do so, the pay practice will be unlawful unless objectively justified. For the definition of what amounts to indirect discrimination, one looks to the SDA 1975. This Act applies only to discrimination in fields other than pay (non-pay cases) but it must define the concept of indirect sex discrimination in other fields as well. As originally enacted, section 1(1)(b) of this Act required a complainant to prove that the employer had imposed a requirement or condition which was such that the proportion of women who could comply with it was considerably smaller than the proportion of men who could comply. In practice this was usually thought to require the female complainant to identify the pools of men and women which were to be compared and to produce a statistical analysis of the effect of the requirement or condition on both pools. As presently enacted, section 1(2)(b) requires the complainant to prove only that the employer has applied a provision, criterion or practice to all but which puts women at a particular disadvantage when compared with men and puts the claimant at that disadvantage. Proving a particular disadvantage on women will not necessarily require the production of statistics although that will often be the way in which the particular disadvantage will be demonstrated. If it is shown that women (and the claimant) are put at a particular disadvantage, the defendant is put to objective justification.

It will be apparent that (unlike section 1(3) of the EqPA) the SDA does not provide an express defence of showing that the difference of treatment is not ‘the difference of sex’. The argument was advanced before us (and in Surtees) that, in a case of indirect discrimination in a non-pay case, once disparate adverse effect or a particular disadvantage has been demonstrated, no such defence is available; it is not open to the defendant to show that the difference of treatment is unrelated to sex. The only possible conclusion is that the difference of treatment is discriminatory and the court or tribunal moves immediately to objective justification. The argument then goes that, if that is the position in a non-pay case, the same situation should and does apply in an equal pay case. It is argued that the decision in Enderby shows that that was the approach of the ECJ in a case of indirect discrimination in the field of pay; the UK courts must follow that.

I do not accept that submission. The position in a non-pay case of alleged indirect discrimination is that the claimant must prove that the provision puts women at a particular disadvantage. While putting the burden of proof squarely on the claimant, the provision leaves open to the defendant the opportunity to demonstrate that what might appear to be a disadvantage to women arises from factors wholly unrelated to gender. So, even though the SDA does not provide an express defence comparable to that under section 1(3) of EqPA, there is an implied one because it is always open to a party who does not bear the burden of proof to shoulder an evidential burden and disprove his opponent’s case. There must be such an implied possibility because the purpose of the legislation is to prevent sex discrimination, including unjustifiable indirect discrimination. A defendant is not to be held to have discriminated - and be put to justification of his practice - merely because it has given rise to a statistical imbalance.

Thus, in my judgment, because the onus of proof under the SDA lies on the claimant and because, as a matter of principle, objective justification is only required if the provision has the effect of discriminating against women, the SDA must implicitly leave it open to a defendant to show that, notwithstanding the evidence which shows that the practice gives rise to disparate adverse impact on women or puts women at a particular disadvantage, that adverse impact or particular disadvantage have nothing whatever to do with the gender of those involved.

In my judgment, if that defence is available in a non-pay case under the SDA, it must also be available in an indirect discrimination case in the field of pay. The principles must be the same for pay and non-pay cases. Indeed, the words of section 1(3) suggest that that defence is available in equal pay cases generally, whether of alleged direct or indirect discrimination. Armstrong says that it is available in cases of alleged indirect discrimination in the field of equal pay. I think that is right. It is true that, in Enderby, the ECJ did not mention its availability but in my view that does not mean that the Court ruled it out, as a theoretical possibility. On the facts of that case, the explanation that was advanced for the pay differential between the two groups was that historically there had been separate negotiating machinery which had not in itself been discriminatory. But, given the history of adverse impact on the female dominated group as compared with the male dominated comparators, the inference that the difference was sex tainted was very strong and the explanation proffered completely failed to demonstrate that the pay differential was wholly unrelated to gender.

My conclusion is that whether the alleged indirect discrimination arises in the field of pay or non-pay, it is always open to a defendant to demonstrate that, notwithstanding the appearance that the practice puts women at a particular disadvantage, in fact the apparent disadvantage has arisen due to factors which are wholly unrelated to gender.

I think that the difference between the statutory provisions in pay and non-pay cases can be understood by considering the effect of the burden of proof. In a non-pay case under the SDA, the burden of proving that the provision puts women (and the claimant) at a particular disadvantage lies squarely on the claimant. So, the claimant must show indirect sex discrimination; the burden then passes to the defendant to justify. Because the burden lies on the claimant, there is no need for an express defence of ‘this is nothing to do with sex’ but that possibility is always there. In an equal pay case, once the individual female claimant has proved that she is paid less than a singular male comparator, the section 1(3) defence is triggered. On the face of the statute, the employer shoulders the whole of the remaining burdens, that is of negating discrimination (direct and indirect) and then of objective justification if he fails to negate indirect discrimination. If that is right, it makes sense that there should be an express defence of ‘nothing to do with sex’ in equal pay cases whereas no such defence is necessary in non-pay cases because the burden of proving discrimination lies on the claimant anyway. Of course, in Nelson v Carillion Services Ltd [2003] ICR 1256, this Court said that, in an equal pay case where indirect discrimination is alleged, the burden of proving that the pay practice disadvantages women lies on the claimant. It is said that that had to be so to bring it in line with the SDA. As this Court said in Surtees, there are powerful arguments for saying that Nelson was wrongly decided. But it is binding unless and until the Supreme Court says otherwise. However, provided that tribunals are prepared to assist claimants with orders for the production of statistics where requested and provided that tribunals recognise that the real question is whether the evidence shows that the pay practice puts women at a particular disadvantage, it will not greatly matter where the burden of proof lies.

Having concluded that it is always open to an employer in a case of alleged indirect discrimination in the field of pay to demonstrate that the particular disadvantage apparently demonstrated is nothing to do with gender, it is important that tribunals examine such a contention with great care. The defence will almost always, as in this case, entail the giving of an historical explanation for how the pay arrangements came to be what they now are and how the complainant (and other women like her) came to be paid less than their comparators. Of course, the tribunal must be satisfied that the explanation is genuine. It was so satisfied in this case. But it must also be satisfied that the difference of treatment is not in any way related to the difference of sex. Where the disadvantaged group is heavily dominated by women and the group of advantaged comparators is heavily dominated by men (as they were in this case), the inference of sex taint will readily be drawn and it will be difficult for the employer to prove its absence. Where the evidence reveals that the work done by the disadvantaged group has historically been done by women and the work done by the advantaged group has historically been done by men, there may well be a basis for inferring that the employer has (in the past) had a subconscious attitude that women do not need to earn as much as men and that the present pay arrangements are a legacy of that attitude. As I understand it, that is what this tribunal meant when they said that there was evidence of stereotypical assumptions. In such a case, the inference that the difference of treatment is sex tainted will readily be drawn.

Yet, in the present case, even though the disadvantaged group of carers was heavily dominated by women and the advantaged groups by men and even though there was evidence of stereotypical assumptions, the tribunal found that there was no sex taint. Their reasoning was that they accepted that, in the 1960s it had been necessary to pay the male gardeners and street cleaners a bonus in order to improve their productivity. It was not necessary to pay the female carers a bonus because they were already working at an acceptable level of productivity. It was not practicable to devise a bonus system related to the productivity of carers. That was the reason why the carers were being paid less than their comparators. It was not because they were women. The tribunal then concluded that the reason was not discriminatory. It was true that there had been no discrimination on the ground of gender but that meant only that there had been no direct discrimination. But direct discrimination was not alleged.

The tribunal’s reasoning did not address the question of whether the pay arrangements, which on their face were gender neutral, unintentionally put women employees at a particular disadvantage when compared with the male employees. They plainly did. The claimants’ disadvantage did not result from the fact that they were women; that would have been direct discrimination. But the disadvantage was indirectly causally linked to their gender. In my view, the history of the payment of bonuses to the street cleaners and gardeners and not to female carers was redolent of gender bias. It was quite impossible to say that the practice was nothing to do with sex. It was plainly indirectly discriminatory and required objective justification.

In my view, where the statistics show that the pay practice has produced an adverse impact on women over a long period and where the statistics are convincing, it will generally be difficult for an employer to show that the adverse impact had nothing to do with sex. In Surtees in the EAT [2007] ICR 1644, Elias P, as he then was, sought for examples of the kind of case in which it might be possible for an employer to explain how a statistical gender imbalance had arisen without there being any sex taint. It appears to me that he found it hard to think of any such circumstances and gave only one example, at paragraph 54 of his judgment. He postulated two groups of workers (A and B), both, in the past, comprising only or mainly men but where group A workers were always paid more than group B. Over the course of time, the composition of group B changed so as to become predominantly female but group A remained predominantly male. If the work of the two groups was rated as equivalent and if the women in group B complained of unequal pay, the statistics would show an adverse impact on them. Elias P said that a tribunal would readily be persuaded that the difference in pay was not sex-tainted notwithstanding the statistical analysis showing an adverse impact on the women in group B. The employer would then establish the section 1(3) defence without the need to justify. I agree with him that that is an example of a situation in which the employer could show the absence of sex taint and would avoid justification even though there was disparate impact. I am not saying, of course, that the circumstances envisaged by Elias P are the only ones in which it will be possible for the employer to demonstrate that the adverse impact was not sex tainted. There will be others although I cannot think of any.

Before leaving the decision in Armstrong, I want to draw attention to an error in Arden LJ’s step-by-step guide to determination of an equal pay case in which the genuine material factor defence is raised. This starts at her paragraph 32 (as quoted by Pill LJ in paragraph 33 above). Arden LJ said that, as the first stage of the process, the claimant must produce a gender-based comparison showing that women doing equivalent work were being paid less than men. With great respect, that is not correct. This is an error which can properly be corrected because it is not an essential part of the reasoning of the decision. It is clear from section 1(2) of the Equal Pay Act 1970 that, in order to reach ‘first base’ and to put the employer to proof of its genuine material factor defence under section 1(3), the claimant need show only that she (as an individual) is being paid less than an individual comparable male worker. She does not have to produce statistics in order to trigger the section 1(3) defence, although if Nelson is correct, she will have to if she wishes to allege indirect rather than direct discrimination. Armstrong was a case of alleged indirect discrimination in which statistics were relied upon which probably explains why Arden LJ overlooked the fact that section 1(3) provides a simpler way for a claimant to get her case ‘off the ground’ than does section 1(2)(b) of the SDA. It appears that Arden LJ mistakenly equated an equal pay case with a non-pay case in which the claimant seeks to prove indirect discrimination by means of statistics showing adverse impact on women. As I have explained above, they are not the same. The point I make is that, to trigger the genuine material factor defence, the claimant need only show that she is being paid less than her comparator. That means that, as well as Arden LJ’s first proposition, her final proposition (the last sentence of her paragraph 34) is also incorrect in that the disparately adverse effect does not denote the trigger at which the rebuttable presumption of sex discrimination arises. The trigger may well be the demonstration that the claimant (singular) is being paid less than a single male comparator. Only where indirect discrimination is alleged does there have to be evidence of the collective position of women.

I, too, would allow the appeal and remit the case to the Employment Tribunal on the issue of justification. 

Subject to the same caveat expressed by Smith LJ in the final paragraph of her judgment, I am convinced that Armstrong is correct.  My reason can be stated succinctly.  An employer who can prove that a difference in pay as between a man and a woman is due to a material factor which is not the difference of sex is protected by the defence contained in section 1(3) of the Equal Pay Act.  In Armstrong, Arden and Buxton LJJ, in the passages referred to by Pill LJ, took that to be axiomatic and consistent with what Lord Nicholls said in Marshall.  I agree.

That said, on any view the employer did not discharge the burden in the present case.  I entirely agree with the reasons given by Pill LJ by reference to the statistical evidence and the history of stereotyping.  Whilst I accept that an employer who seeks to avail himself of the section 1(3)/ Armstrong defence will generally have a difficult task when faced with ostensibly significant statistics, I prefer to resist the language of exceptionality and the temptation to hypothetical creativity.