England and Wales Court of Appeal (Civil Division) Decisions

IN THE SUPREME COURT OF JUDICATURE
IN THE COURT OF APPEAL (CIVIL DIVISION)
ON APPEAL FROM THE EMPLOYMENT APPEAL TRIBUNAL

		Royal Courts of Justice Strand London W2A 2LL
		21st May 1998

Lord Justice Potter:

INTRODUCTION

1. This is an appeal by London Underground Limited ("LU") against a decision of the Employment Appeals Tribunal dated 13th January 1997 whereby it dismissed an appeal by LU against a decision of the Industrial Tribunal at London (North) dated 17th November 1995 that LU had unlawfully discriminated against Miss Susan Edwards ("the applicant"), one of their former employees, on the grounds of her sex contrary to the Sex Discrimination Act 1975 ("S.D.A."). The discrimination complained of and found to have been established was that, in applying a condition or requirement that the applicant enter into a new contract and work new rostering arrangements, LU made it impossible for her to continue in her employment.
2. The decision of the Industrial Tribunal and the judgment of the E.A.T. the subject of this appeal were respectively the second decision and judgment in the course of proceedings between the applicant and LU.
3. By its original decision the Industrial Tribunal upheld the applicant's complaint. However, in doing so, when considering the question under S.D.A. s.1(1)(b)(i) whether the proportion of women who could comply with the requirement was considerably smaller than the proportion of men who could comply, the Industrial Tribunal selected the wrong "pool" of employees for comparison, namely those employees who were single parents. On appeal, the E.A.T. held that the Industrial Tribunal should have based its decision upon a pool consisting of all train operators to whom the new rostering arrangements applied; it remitted the case for hearing before another Industrial Tribunal, see London Underground Ltd -v- Edwards 1995 IRLR 355 ("Edwards No. 1").

THE FACTS

4. Miss Edwards first entered the employment of LU in September 1983. She qualified as a train driver (called a train operator) in 1987 and in the same year she became the mother of a child of whom she had sole care. For most of her employment she worked for LU at the Northfields train depot where rostering arrangements were in force until the end of 1992 which were based on a system of three shifts per twenty-four hours. Staff were allocated to a shift under the roster well in advance so that, if they wished, they were able to swop shifts. Because extra payments were involved for work during unsocial hours, in order to accommodate her domestic and child care arrangements, the applicant was able, by making the necessary swops, to organise for herself a shift pattern in which she could work from 0800 to 1600 hours or 0830 to 1630 hours during the weekdays and a single shift on alternate Sundays. The tribunal found that:

"As a result she was able, as single parent, to do her job and be at home to look after her child. She however suffered a financial penalty in working these shifts as these shifts were the ones that did not attract a bonus payment [an unsocial hours payment]."

5. In 1991 LU brought into effect a company re-organisation plan with the aim of reducing costs and manpower and increasing efficiency. This involved a new system of rostering. Under the old system some of the shifts were very short and it appeared to LU that this fact, coupled with the increased use of advanced technology which reduced the need for guards on trains meant that it was possible to reduce manning levels and make considerable financial saving. At the same time, a voluntary severance scheme was introduced, coupled with the offer of enhanced pension payments, so that compulsory redundancies were avoided.
6. Under the new system, flexible shifts were to be introduced with a minimum of 4 hours and a maximum of 8¾ hours, the duties beginning at 4.45 am and ending at 1.30 pm, with Sundays part of the working week. Operators were expected to work over a seven-day period and to be paid a fixed salary, working an average of 38.5 hours per week over a period of four weeks. Nonetheless, the working of the shift pattern might involve an operator working on eleven consecutive days. As a result of the scheme, a person wishing to work social hours during the daytime would have to exchange shorter shifts for longer shifts with no extra remuneration to attract a person to work the unsocial hours.
7. LU had contemplated introducing a scheme called the "Single Parent Link" which would involve isolating eleven separate shifts, at either end on the line, from the general roster. It was proposed that those who wished to could apply to work those shifts rather than go into the general roster. However, this proposal was rejected by the Trade Unions, who favoured a creche arrangement, and the proposed scheme was abandoned in early September 1992.
8. When the new system was introduced the applicant made it clear to LU that if she could not be satisfied that her ability to exchange her shifts could continue she would not be able to carry on in LU's employment. She and one other female operator had made an application to join the Single Parent Link Scheme. However following its abandonment, when the applicant was given a letter in November 1992 giving her details of the changes in her contract, she made it clear that she did not agree with the proposed changes and that she was not prepared to sign the new contract unless assured she would not have to work the new rosters as proposed. The tribunal found as a fact that it was necessary for the applicant to work during the day because she had sole care of the child and that under the new system it would have been more difficult for her to arrange any exchange of shifts and it would have meant that she would have had to work longer hours than previously. She was presented with the alternative of either signing an acceptance of the new roster or facing dismissal. As a compromise, the parties agreed that she should sign her new form of contract and then be given a voluntary severance payment. In December 1992 she received such a payment and her employment was terminated.

THE INDUSTRIAL TRIBUNAL'S DECISION

9. The applicant's complaint was that she had been indirectly discriminated against on the grounds of her sex contrary to S.D.A. Section 1(1)(b), as amended, which provides that a person discriminates against a woman if:

"(b) he applies to her a requirement or condition which applies or would apply equally to a man but-

(i) which is such that the proportion of women who can comply with it is considerably smaller than the proportion of men who can comply with it; and

(ii) which he cannot show to be justifiable irrespective of the sex of the person to whom it is applied; and

(iii) which is to her detriment because she cannot comply with it."

10. The Industrial Tribunal correctly directed themselves as to the four questions which they had to consider;

(a) What was the relevant "requirement or condition" which was applied to the applicant?

(b) Could she comply with it?

(c) If not, was it one which "a considerably smaller proportion of female train operators than of male operators could comply"?

(d) If so, was it justifiable.

The tribunal's answer to Question (a) was: "The new rostering arrangements in the new contract ... [which] .. affected the applicant and .. ultimately led to the termination of her contract".

Their answer to the second question was: "No". They held:

"It is clear from the evidence that no reassurance was given to her, and we are not satisfied from the evidence that we heard that the swopping arrangements that would have enabled her to continue with her work would continue in the way that were [sic] satisfactory. As a single parent she was torn between the need to do her job and the need to care for her child and these new rostering arrangements in the new company plan did not satisfy her needs."

The tribunal's answer to the third question was that the requirement was one with which a considerably smaller proportion of the female than male train operators could comply. I shall return to the findings in support of that answer in detail below.

In answer to the fourth question the tribunal found that the requirement imposed on the applicant was not justifiable on the grounds that LU:

"could have easily, without losing the objectives of their plan and re-organisation, have accommodated the Applicant who was a long-serving employee. They were aware of her particular difficulties quite early on and, after the failure of the Single Parent Link in September 1991, she had set out her misgivings and her difficulties in writing to the management. They did not address themselves to these issues and therefore we find that they have not justified this act of discrimination."

THE DECISION OF THE EMPLOYMENT APPEAL TRIBUNAL ("E.A.T.")

11. On appeal to the E.A.T., LU did not seek to argue that the Industrial Tribunal erred in law in its answers given to the first two questions. However, it was submitted that they had erred in law in their answers to the third and fourth questions. That appeal was dismissed. Upon appeal to this court, LU no longer challenge the answer to the fourth question. Their appeal relates solely to the answer to the third question, which for convenience I shall refer to as "the disparate impact question": cp. the phrase "disparate adverse impact" in R -v- Secretary of State for Employment ex parte Seymour-Smith [1995] ICR 859 at 942 per Neill L.J.

THE DISPARATE IMPACT QUESTION

12. It is contended by LU that, having regard to the findings of fact made by the Industrial Tribunal as to the proportion of male and female operators who were able to comply with the requirement to work to the new rostering requirements, the Industrial Tribunal was wrong in law in the conclusion it reached that the proportion of women who could comply was considerable smaller than the proportion of men who could comply. It is thus necessary to set out in extenso the findings of fact which underlay that answer at paragraphs 9 and 10 of the decision of the Industrial Tribunal. They read as follows:

"9. The third requirement is whether it was a requirement or condition with which a considerably smaller proportion of female train operators could comply as compared with male train operators. We have heard from the evidence before us, and it was accepted by Mr Bean, that all the 2,023 male operators complied with this condition of the new rostering. We heard evidence from the Applicant that out of 21 female train operators she was the only one who positively complained that she could not comply with the new rostering arrangement. However, the Respondent's witnesses confirmed that there was a Mrs Quinlan who had also applied for the Single Parent Link arrangement when it was proposed but she had signed the contract and had worked for two or three weeks under the new rostering system. It was accepted that she was then moved because it was said that she had a bad back as she was pregnant and eventually she took a career break. She is still apparently employed by the Respondents but she was not called as a witness by either the Respondents or the Applicant and we can only conclude from the evidence that we heard that, out of 21 female train operators, two of them did not continue with the new rostering a month after it came into operation. The Respondents have argued that the Applicant should have tried the new rostering arrangement to see if it could have worked out. However, an employee faced with a choice between working a new roster and, if it did not prove satisfactory, having to leave or being dismissed for breach of contract, and the alternative of an enhanced voluntary severance package, could reasonably not risk working the roster when they were quite clear in their own mind that they would not be able to do so for very long.

10. We now have to consider whether the one or possibly two train drivers out of 21 female train drivers as compared with 2,023 male train operators can be considered a "considerably smaller" proportion of the whole which we consider are the train operators as a whole. We were referred by Mr Bean to Staffordshire County Council -v- Black [1995] IRLR, where it was held that a proportion of 89.5 per cent of females compared with 97 per cent of male teachers was not a "considerably smaller" proportion within the meaning of Section 10(1)(b) of the Sex Discrimination Act. Mr Bean has argued that this proportion in this case is less than a 5 per cent difference and therefore was not a big enough proportion to come into this category and to be classified as "considerably smaller". It is clearly established from the case law that this is a question of fact for the Tribunal to decide, and we find that taking into account the number of male train operators as compared to the very few female train operators (2,023 to 21) and the fact that only one person, the applicant, could not comply and also added to the fact that it is common knowledge that females are more likely to be single parents and caring for a child than males, it is clear that this was a condition or requirement that a "considerably smaller" number of females could comply with."

13. Mr Bean, Q.C. for LU, has attacked both the approach and the conclusion of the Industrial Tribunal as well as the reasoning of the E.A.T. in upholding it. Put simply, his submissions were as follows:

(1) It was common ground that for the purposes of comparing the proportion of women and men employees who could comply with the new rostering arrangements, the "pool" was that of all train operators to whom the new rostering arrangements applied (see: Edwards No.1) which (as counsel were agreed) consisted of, and thus was limited to, those in the employment of LU at the time the arrangements were introduced

(2) The Industrial Tribunal found that the pool consisted of the 2023 male train operators, of which all, i.e. 100%, could comply, and twenty-one female train operators of whom all but 1 (Mrs Edwards), i.e. 95.2%, could comply.

(3) That being so, the question for the Industrial Tribunal was, and was no more than, whether or not 95.2% was a "considerably smaller" proportion than 100%.

(4) In finding that it was a considerably smaller proportion, the Industrial Tribunal erred in law. While Mr Bean acknowledged that, when dealing with matters of fact and degree in the field of industrial practice or discrimination, the decision of an Industrial Tribunal is normally to be regarded as that of an industrial jury, he nevertheless submitted that the question whether a proportion is "considerably smaller" in this context is to an extent a question of statutory interpretation and to an extent a "threshold" matter as to which the Court of Appeal should be prepared to intervene in an appropriate case. In this respect he referred to certain authorities to which I shall turn below.

(5) Mr Bean further submitted that, even if the decision was essentially one of degree for the Industrial Tribunal, in reaching it the tribunal had taken into account matters which were strictly irrelevant given the circumscribed nature of the task to be performed. In particular,

(i) they attached significance to "the number of male train operators as compared to the very few female operators (2023 to 21)". Mr Bean said that was irrelevant to the issue for decision, namely the comparative proportion of women train operators able to comply;

(ii) they did not limit themselves to making the straight percentage comparison which arose on the figures revealed by the pool, but acted also on:

"The fact that it is common knowledge that females are more likely to be single parents and caring for a child than males."

This additional and, as Mr Bean again submitted, irrelevant consideration appeared to have driven the tribunal from its finding potentially favourable to LU that

"Only one person, the Applicant, could not comply"

to the conclusion that

"This was a condition or requirement that a "considerably smaller" number of females could comply with."

14. The authority principally relied on by Mr Bean under his submission (4) above, was the Seymour-Smith case in which, on appeal to the House of Lords, the House has referred a number of questions to the European Court of Justice. The third question (see [1997] ICR 371 at 381 B) is:

"What is the legal test for establishing whenever a measure adopted by a member state has such a degree of disparate effect as between men and women as to amount to indirect discrimination for the purposes of Article 119 of the E.C. Treaty unless shown to be based upon objectively justified factors other than sex?"

15. In Seymour-Smith, in relation to a submission by counsel for the Secretary of State as to the point at which an apparent disparity requires to be justified, Neill LJ in giving the judgment of the Court of Appeal, observed of various relevant decisions of the European Court of Justice set out at p.949-950 that:

"In every case the Court of Justice referred not to a simple difference between the number or percentage of women and the number or percentage of men within a particular group but to "a considerably smaller or larger percentage" or to "a much larger or lower number". Counsel for the parties provided us with a useful summary in the form of an agreed note comparing the French and English text of the judgments in the eight cases to which we have just referred and also in Enderby -v- Frenchay Health Authority (Case C-127/92) [1994] ICR 112. It is noteworthy that in a majority of these cases the French term which is used to describe the disparity is "Un nombre beaucoup plus élevé:" ...[authority cited]...

In these circumstances we would accept the submission made on behalf of the Secretary of State that before a presumption of indirect discrimination on the ground of sex arises there must be a considerable difference in the number or percentage of one sex in the advantaged or disadvantaged group as against the other sex and not simply a difference which is more than de minimis."

16. Similarly, in the allied field of equal pay, in Barry -v- Midland Bank PLC [1998) IRLR 138, Peter Gibson LJ, giving the judgment of this court (at 143 Para. 30) observed:

"but the consistent approach of the European Court has been to go beyond the fact that a measure is formulated in neutral non-discriminatory terms and to see whether the measure works to the disadvantage of far more women than men, and if so, to leave it the national court to determine in the light of all the circumstances whether such measure is justified by objective reasons unrelated to any discrimination on grounds of sex." (emphasis added)

Mr Bean also referred us to R -v- Secretary of State ex p. Unison [1996] ICR 1003 in which the Divisional Court had to consider the proportions of male and female employees who enjoyed the two-year qualification for unfair dismissal claims. On the most recent statistics the gap was shown to be 4%, though the evidence was not complete. Otton LJ, with whom Newman J agreed, stated:

"without making a final decision ... I would on the evidence currently available be inclined to the view that if there is now only a four per cent disparity then this would fall within the de minimis exception. Thus, if the Secretary of State had considered, or if he were to consider this question, it would be open to him to conclude that the disparity was less than considerable. There would thus be no obligation upon him to consider objective justification on this occasion."

Thus, submitted Mr Bean, the meaning of the words "considerably smaller" in S1(1))b) properly interpreted in accordance with the views of the Court of Appeal expressed above, should have led the Industrial Tribunal to conclude that discrimination had not been made out in the applicant's case.

17. Mr Allen QC for the applicant on the other hand argued as follows. He accepted stages (1), (2) and (3) of Mr Bean's argument, but resisted the proposition that the Industrial Tribunal erred in law: see stage (4). He accepted that the approach of the cases in general terms was that there must be a "considerable difference" and not merely a difference properly described as de minimis, but urged that, within the broad and uncertain confines of the phrase, this court should be slow to interfere with the decision of the Industrial Tribunal which had heard and considered the evidence and applied it in its industrial setting.
18. Mr Allen submitted that the approach to the disparate impact question should not be formulaic at any stage. It was in principle wrong to draw a line, or to nominate a specific percentage difference, above or below which an allegation of discrimination succeeded or failed. Thus it must be of limited use to make comparisons with figures in other cases in which the subject matter may be different (e.g. cases involving the workforce of an entire industry or a particular section of the national population), and in which the relevant requirements or conditions being imposed by the employer upon a mixed workforce may have highly variable effects. In relation to a given employer or area of activity, careful analysis and consideration may be necessary because a number of factors may operate to produce a small but nonetheless industrially significant difference in terms of discrimination.
19. Thus, in any given case (and Mr Allen submitted this is one)

(i) there may be historical reasons (discriminatory or other) leading to a small number of women (or men) in the workplace;

(ii) there may be small overall numbers involved in the pool under consideration, so that a variant of one employee in either compared group within the pool may have a large effect on any percentage difference;

(iii) there may be some fortuitous circumstance affecting the numbers within the pool or within either compared group, or affecting the ability of particular group members to comply with a condition or requirement operative at the time of the comparison but atypical of the usual position;

(iv) in such cases, to convert the relevant proportions into percentages can be misleading if the background, numbers and makeup of the compared groups are not considered and taken into account so as to inform the decision as to whether or not the difference is "considerably smaller".

20. In relation to the matters stigmatised as irrelevant by Mr Bean, Mr Allen submitted that

(i) The Industrial Tribunal was right to have regard to the large discrepancy in numbers between male and female operators. The figure of 100 men for every 1 woman indicated that it was either difficult or unattractive for women to work as train operators. Thus a condition which made it even more difficult for women to perform or continue in such work was of particular significance.

(ii) They were also entitled to take into account their own knowledge and experience that the burden of child care falls upon many more women than men and that a far greater proportion of single parents with care of children are women rather than men. Consequently the fact that one woman only was shown positively not to have been able to comply with LU's rostering requirements did not reflect a fortuitous error and indeed was likely to represent a minimum impact on the female workforce.

(iii) in relation to (ii) above, Mr Allen pointed out that, while the Industrial Tribunal did not find as a fact that Mrs Quinlan, the second woman who applied for the Single Link Arrangement, could not comply with the new arrangement, their reference to her was such that their finding was effectively that the percentage difference was at least 100: 95.2 i.e. if anything it was an under-estimate of the position, which by implication tended to confirm their view that the number and proportion of women able to comply would always be substantially smaller bearing in mind the known disparity in numbers between female and male single parents.

21. Finally, Mr Allen submitted that, in the light of his above submissions, the argument of Mr Bean was in reality an assertion that the decision of the Industrial Tribunal constituted a perverse finding in the course of an exercise which was essentially a factual one, involving as it did the assessment of the validity and significance of the particular figures and their effects in a particular industrial context. He pointed out that courts have fought shy of putting any specific numbers or limits upon the expression "considerably smaller" and submitted that they have been right to do so, though for good measure he referred to our attention the decision in Greater Manchester Police Authority -v- Lea [1990] IRLR 375. In that case, in relation to a pool consisting of the commercially active population of the United Kingdom, the E.A.T. upheld a decision of an Industrial Tribunal that the proportion of 95.3% of men unable to comply with a particular requirement was "considerably smaller" than the proportion of women, namely 99.4% (a difference of 4.1%). In that case, Knox J observed (at p.375):

"The way in which the case was put to us by Mr Tattersall at one stage in his argument was that if one had gone into the street and asked people whether 95.3% was a considerably smaller proportion than 99.4% there would not have been many takers for that view. That formulation in our judgment lets a rather significant cat out of the bag because it does connote that some takers might have been found for that view. It is of course wholly undesirable that one should determine the case upon the particular formulation that counsel relies upon but in our judgment it is illustrative of the heart of this particular problem which is whether or not the assessment that the Industrial Tribunal has reached on this aspect of the matter is one which any reasonable tribunal could have reached. We are not expressing the view that had the question been one for us we would have reached the same conclusion. That is not the function of this tribunal. It is of course notorious that this tribunal is solely concerned with questions of law and the question of law is not 'do we agree with what the Industrial Tribunal said on this issue' but 'do we consider that no reasonable tribunal could have reached its conclusion'. The latter question we unhesitatingly answer 'No, we are not satisfied that no reasonable tribunal could have reached this conclusion', and it would be inappropriate for us to express the view that we - or perhaps it might be the majority of us - would have reached on the question had it been one for us to answer."

CONCLUSIONS

22. I accept the submissions of Mr Allen. In my view there is a dual statutory purpose underlying the provisions of S1(1)(b) and in particular the necessity under sub-paragraph (i) to show that the proportion of women who can comply with a given requirement or condition is "considerably smaller" than the proportion of men who can comply with it. The first is to prescribe as the threshold for intervention a situation in which there exists a substantial and not merely marginal discriminatory effect (disparate impact) as between men and women, so that it can be clearly demonstrated that a prima facie case of (indirect) discrimination exists, sufficient to require the employer to justify the application of the condition or requirement in question: see sub-paragraph (ii). The second is to ensure that a tribunal charged with deciding whether or not the requirement is discriminatory may be confident that its disparate impact is inherent in the application of the requirement or condition and is not simply the product of unreliable statistics or fortuitous circumstance. Since the disparate impact question will require to be resolved in an infinite number of different employment situations, well but by no means comprehensively exemplified in the arguments of Mr Allen, an area of flexibility (or margin of appreciation), is necessarily applicable to the question of whether a particular percentage is to be regarded as "substantially smaller" in any given case.
23. The first or preliminary matter to be considered by the tribunal is the identification of the appropriate pool within which the exercise of comparison is to be performed. Selection of the wrong pool will invalidate the exercise, see for instance Edwards No.1 and University of Manchester -v- Jones [1993] ICR 474, and c.p. the judgment of Stephenson LJ in Perera -v- Civil Service (No.2) [1983] ICR 428 at 437 in the context of racial discrimination. The identity of the appropriate pool will depend upon identifying that sector of the relevant workforce which is affected or potentially affected by the application of the particular requirement or condition in question and the context or circumstances in which it is sought to be applied. In this case, the pool was all those members of the LU workforce, namely train operators, to whom the new rostering arrangements were to be applied (see paragraph 3 above). It did not include all LU employees. Nor did the pool extend to include the wider field of potential new applicants to LU for a job as a train operator. That is because the discrimination complained of was the requirement for existing employees to enter into a new contract embodying the rostering arrangement; it was not a complaint brought by an applicant from outside complaining about the terms of the job applied for. There has been no dispute between the parties to this appeal on that score. However, Mr Bean has placed emphasis on the restricted nature of the pool when asserting that the Industrial Tribunal were not entitled to look outside it in any respect. Thus he submitted they should not have taken into account, as it apparently did, its own knowledge and experience, or the broad national "statistic" that the ratio of single parents having care of a child is some 10:1 as between women and men.
24. In my view Mr Bean was incorrect in that last respect. An Industrial Tribunal does not sit in blinkers. Its members are selected in order to have a degree of knowledge and expertise in the industrial field generally. The high preponderance of single mothers having care of a child is a matter of common knowledge. Even if the "statistic" i.e. the precise ratio referred to is less well known, it was in any event apparently discussed at the hearing before the Industrial Tribunal without doubt or reservation on either side. It thus seems clear to me that, when considering as a basis for their decision the reliability of the figures with which they were presented, the Industrial Tribunal were entitled to take the view that the percentage difference represented a minimum rather than a maximum so far as discriminatory effect was concerned.
25. Equally, I consider that the Industrial Tribunal was entitled to have regard to the large discrepancy in numbers between male and female operators making up the pool for its consideration. Not one of the male component of just over 2,000 men was unable to comply with the rostering arrangements. On the other hand, one woman could not comply out of the female component of only 21. It seems to me that the comparatively small size of the female component indicated, again without the need for specific evidence, both that it was either difficult or unattractive for women to work as train operators in any event and that the figure of 95.2% of women unable to comply was likely to be a minimum rather than a maximum figure. Further, if for any reason fortuitous error was present or comprehensive evidence lacking, an unallowed for increase of no more than one in the women unable to comply would produce an effective figure of some 10% as against the nil figure in respect of men; on the other hand, one male employee unable to comply would scarcely alter the proportional difference at all. Again, I consider Mr Allen is right to point out in relation to Mrs Quinlan that, albeit the Industrial Tribunal lacked the evidence to find as a fact that she could not comply, the reference to her indicates that they had her uncertain position in mind when assessing the firmness of the figure of only 4.8% as the basis for a finding of prima facie discrimination.
26. Finally, although the Industrial Tribunal did not state that they placed reliance upon the LU's original proposal for the Single Parent Link Scheme, which was designed to deal with precisely the kind of difficulties which faced the applicant, they might legitimately have done so in my view; it was a recognition by LU of the likely adverse impact of the proposed rostering arrangements on those members of the workforce who were single parents and thus, afforded support for the Industrial Tribunal to take the view that the effect of the rostering arrangements, unalleviated by such a scheme, was discriminatory in effect.
27. Thus I do not consider that the decision or reasoning of the Industrial Tribunal is open to attack unless Mr Bean can make good his main argument that their conclusion was contrary to the wording of the statute, or alternatively a decision which no reasonable tribunal could have reached. As to the first, in the light of the previous expressions of view of this court to which I have referred I have not found this an easy question. However, I have concluded that the argument must fail in the circumstances of this case.
28. In many respects, no doubt, it would be useful to lay down in relation to S.1(1)(b) a rule of thumb or to draw a line defining the margin within, or threshold beyond which, in relation to small percentage differences, the lower percentage should not reasonably be regarded as "considerably smaller" than the higher percentage. However, it does not seem to me appropriate to do so. For the various reasons discussed in this judgment, and because of the wide field and variety of situations in which the provisions of the Section are to be applied, the circumstances and arguments before the adjudicating tribunal are bound to differ as to what in a particular case amounts to a proportion which is "considerably smaller" for the purposes of determining the discriminatory or potentially discriminatory nature of a particular requirement or condition. If a figure were to be selected in the field of employment, it would be likely to vary according to the context, and in particular as between a case where the requirement or condition is applied on a national scale in respect of which reliable supporting statistics are available and those where it is applied in relation to a small firm or an unbalanced workforce where the decision may have to be made on far less certain evidence and to a large degree upon the basis of the Industrial Tribunal's own experience and assessment as applied to such figures as are available. The difficulties are well illustrated by this case.
29. Plainly, a percentage difference of no more than 5% or thereabouts is inherently likely to lead an Industrial Tribunal to the conclusion that the requirements of Section 1(1)(b) have not been made out, but I am not prepared to say that such a conclusion must inevitably follow in every case. Nor, for essentially the same reasons as those expressed by Knox J in Greater Manchester Police Authority -v- Lea, am I prepared to say that such a conclusion necessarily followed in this case.
30. I would dismiss this appeal.

Lord Justice Swinton Thomas:

In company with Simon Brown LJ, I have found this a difficult case. At first blush it might seem that the fact that one woman in twenty-one (95.2%) could not comply with the requirement as to rostering arrangements whereas all 2,023 men (100%) could comply would not result in the requirement being such that the proportion of women who could comply was considerably smaller than the proportion of men. However, having considered the submissions made by Mr Bean and Mr Allen with care I have come to the conclusion that Mr Allen's submissions are to be preferred, and that the decision of the Employment Appeal Tribunal is correct. Having had the advantage of reading the judgments of Simon Brown LJ and Potter LJ I agree with their reasoning and conclusions, and there is nothing which I can usefully add to them.

Lord Justice Simon Brown:

The law, the facts and the arguments are fully set out in Potter LJ's judgment and for the most part I do not repeat them. Certain matters, however, are central to the appeal and these for convenience I do set out.

S.1(1) of the Sex Discrimination Act 1975 provides:

"(1) A person discriminates against a woman in any circumstances .relevant for the purposes of any provision of this Act if -

...

(b) he applies to her a requirement or condition which he applies or would apply equally to a man but -

> (i) which is such that the proportion of women who can comply with it is considerably smaller than the proportion of men who can comply with it, and

(ii) which he cannot show to be justifiable irrespective of the sex of the person to whom it is applied, and

(iii) which is to her detriment because she cannot comply with it.

S.5(3) of the Act provides:

"(3) A comparison of the cases of persons of different sex ... under s.1(1) ... must be such that the relevant circumstances in the one case are the same, or not materially different, in the other."

It is s.5(3) which requires that a "pool" of persons be decided upon for the purpose of making the necessary comparison under s.1(1). That pool in the present case is agreed to consist of all the train operators in the appellant's employment at the material time.

It is not in dispute on this appeal (a) that all members of the pool were required to work the new roster, (b) that that requirement was to the respondent's detriment because, as a single parent carer, she was unable to comply with it, and (c) that the appellants could not show the requirement to be justifiable in the respondent's case, given that she had been working for them for nearly ten years. (Had the pool been any larger, one should note, the appellants would in all likelihood have been able to justify the requirement.)

What is now in dispute is whether the requirement was "such that the proportion of women who [could] comply with it [was] considerably smaller than the proportion of men who [could] comply with it", it being clearly established on the authorities that "considerably smaller" in this context means much smaller and not simply smaller to an extent which is more than de minimis.

The essential facts relevant to the determination of that issue were these:

31. The pool comprised 2023 male train drivers and 21 female ones.
32. Only the respondent could not comply with the requirement to work the new roster. In other words, 20 out of the 21 women (95.2%) could comply; all 2023 men could comply.
33. There are in Great Britain 10 lone mothers for each lone father.

The critical question arising on the appeal is this: On these facts can it properly be said that 95.2% is a "considerably smaller" proportion than 100% within the meaning of s.1(1)(b)(i) of the Act?

Contending for the answer No, Mr Bean QC submits in effect that the question answers itself: a difference of under 5% is simply not sufficient. It would, he says, be perverse to describe 95.2% as "considerably smaller" than 100%. On this approach, of course, the fact that only one in twenty-one women was unable to comply with the new roster requirement was of itself fatal to the respondent's cause: it mattered not whether any of the men were affected, nor what proportions of men and women are lone parents in the population as a whole. Those further facts, submits Mr Bean, are simply irrelevant.

Mr Allen QC for the respondent advances two arguments to the contrary. First he submits that the proportion of women to men lone parents in the overall population is relevant to ensuring that the 5% figure here is not fortuitously small. That, he suggests, is always a risk where very small numbers are involved. To my mind, however, this argument carries him nowhere. There is no evidence before us as to what proportion of the working population as a whole are single parents who would be unable to comply with the appellants' roster system and, even if there were, that is not the group which it is agreed must be used here to make the necessary comparisons. In any event the 10 to 1 proportion overall indicates nothing as to whether the 5% figure is fortuitously small or fortuitously large. Given the 10 to 1 proportion overall, one might have expected 0.5% of the men to be similarly affected - i.e. 100 odd out of the 2023 men in the group whereas in fact there were none. But no one suggests that the 5% figure for women was fortuitously high. To my mind the sole relevance of the 10 to 1 statistic here is to preclude the appellants from contending that it is fortuitous to find the women in the group to any degree disadvantaged. It would, for example, have been highly relevant had by chance a man in the group been affected but no woman, and the man then claimed. As it is, and given that the appellants do not contend that there is anything fortuitous about the figures here, the 10 to 1 statistic seems to me of essentially neutral effect.

Mr Allen's second and principal argument is that, even assuming the 5% figure to be correct (as I think we must), i.e. that 95% of the women in the group could comply with the new roster requirement, that proportion could properly be held to be considerably smaller than the 100% proportion of men. It is not to be overlooked, he submits, that at the other end of the scale 5% of the women were disadvantaged as against 0% of the men, a very considerable difference indeed. In the final analysis, he argues, equality of treatment is the paramount consideration.

I confess to having found the point a very difficult one and to having changed my mind more than once during the course of the appeal. The strength of Mr Bean's argument lies in the focus given by s.1(1)(b)(i) to those who can comply, rather than those who cannot. Its weakness, however, to my mind lies in its assumption that Parliament was uninterested in the extent of disparity as between men and women unless only a considerable proportion of the women in the group were found to be disadvantaged by the requirement or condition in the first place.

None of the authorities cited to us seemed to me to bear directly on this point. Certainly none of them appears to decide clearly that, unless a considerable proportion of the women in the group are unable to comply with the requirement in question, it matters not how that requirement affects the men in the group. Sometimes in the judgments phrases are used which might be thought to support one approach rather than the other; for example, in University of Manchester v Jones [1993] ICR 474 at 495 Ralph Gibson LJ stated:

"Parliament enacted that indirect discrimination against women could be found to exist if the proportion of women who could comply with the requirement is considerably smaller than the proportion of men who could comply: it did not put the ground of proof on the proportions within a particular group of people who could not comply with the requirement."

Conversely, in R v Secretary of State for Employment ex parte Seymour-Smith [1995] ICR 889 at 950, Neill LJ giving the judgment of the court said:

"In these circumstances we would accept the submission made on behalf of the Secretary of State that before a presumption of indirect discrimination on the ground of sex arises there must be a considerable difference in the number or percentage of one sex in the advantaged or disadvantaged group as against the other sex and not simply a difference which is more than de minimis." (emphasis added)

I repeat, none of the cases appears to decide the present point.

As to how percentages of this sort have been regarded in the past, one finds only limited assistance, and indeed some inconsistency, in the cases. Whereas, for example, in Staffordshire County Council v Black [1995] IRLR 234, the EAT (presided over, be it noted, by Morison J) upheld the industrial tribunal's finding that 89.5% was not "considerably smaller" than 97% - stating that "the figures speak for themselves", in Greater Manchester Police Authority v Lea [1990] IRLR 372, the EAT (presided over by Knox J) upheld the industrial tribunal's finding that 95.3% was "considerably smaller" than 99.4%.

I can state my conclusions really quite shortly. Given that this legislation is concerned essentially to contrast the impact of a given requirement or condition as between men and women rather than as between the women in the group, it would seem to me wrong to ignore entirely the striking fact here that not a single man was disadvantaged by this requirement despite the vast preponderance of men within the group. Looked at in the round, this requirement clearly bore disproportionately as between men and women, even though only one woman was affected by it. Had there been an equal number of women drivers to male drivers and the same 5% proportion of them been affected, i.e. 100, Mr Bean's argument would remain the same, namely that too large a proportion of women were able to comply with the requirement to leave room for a finding that such proportion was "considerably smaller" than the proportion of men who could comply. It is not an argument I am ultimately prepared to accept. The approach to s.1(1)(b)(i) must, I conclude, be more flexible than this argument allows. Parliament has not, be it noted, chosen to stipulate, as it could, just what difference in proportions would be sufficient. Once, then, one departs from the purely mechanistic approach contended for by the appellants, and has regard to other facts besides merely a comparison between 95% and 100%, the respondent's argument becomes compelling: no other fact could be more relevant than that, whereas 5% of the women were disadvantaged, not one of the 2023 men was. That further consideration in my judgment supports the industrial tribunal's finding here.

I too, therefore, would dismiss this appeal.

ORDER: Appeal dismissed with costs; application for leave to appeal to the House of Lords refused.

(Order not part of approved judgment)