Judgments - Secretary of State for Trade and Industry (Respondent) v. Rutherford and another (FC) (Appellants) and others

(back to preceding text)

    54.  When the case returned to this House [2000] ICR 244, it had to apply the guidance given by the Court of Justice. The House had before it statistics for nine consecutive years, 1985-1993, which showed a modest but persistent and relatively constant differential against women in the effect of the requirement for a two-year qualifying period. The full table is set out at p.252 of the report (and again, in a different and shorter form, at p.258). In terms of the percentage of the total workforce who had completed the two-year qualifying period (the test indicated in para 60 of the judgment of the Court of Justice) the largest differential was in 1986: 77.2% of men to 68.4% of women, that is roughly seven men to six women. The smallest differential was in 1993: 78.4% men and 74.1% women, or roughly sixteen men for fifteen women. These figures were (as my noble and learned friend Lord Nicholls of Birkenhead said at p.258) "in borderline country." They were also in borderline country if one were to look at the corresponding reciprocal figures for men and women who had not completed the qualifying period: 22.9% against 31.6% (about 7 to 10) for 1986, 21.6% against 25.9% (about 11 to 13) for 1993.

    55.  Lord Nicholls referred to these two possible approaches in an important passage of his speech at p.259. After setting out most of para 59 of the judgment of the Court of Justice he observed,

    "This statement appears to envisage that two comparisons should be made: a comparison of the proportions of men and women able to satisfy the requirement ('the qualifiers'), and a comparison of the proportions of men and women unable to satisfy the requirement ('the non-qualifiers'). Thereafter in its judgment the Court considered only the proportions of men and women who were qualifiers.

    Some of the ramifications involved in looking at the composition of the disadvantaged group, as well as the composition of the advantaged group, were explored by the Divisional Court and the Court of Appeal in the present case. Suffice to say, I do not understand the Court of Justice to have rejected use of the figures relating to the non-qualifiers in a suitable case. Indeed the Court has looked at the composition of the disadvantaged group in several cases, although in none of them was there an issue on this point: see for instance, Bilka-Kaufhaus Gmbh v von Hartz (Case C-170/84) [1987] ICR 110, 125, para 31; Nimz v Freie und Hansestadt Hamburg (Case C-184/89) [1991] ECR I-297, 319, para 12, and Kowalska v Freie und Hansestadt Hamburg (Case C-33/89) [1992] ICR 29, 34-35, paras 13-16. Having regard to the conclusion I have expressed above on the issue of disparate impact, it is unnecessary to reach a firm conclusion on this point. I prefer to leave this question open for another occasion."

    Lord Goff of Chieveley and Lord Jauncey of Tullichettle agreed with Lord Nicholls. The other members of the House (Lord Slynn of Hadley and Lord Steyn) disagreed as to the result of the application of the Court of Justice's test to the statistics, but Lord Slynn (with whom Lord Steyn agreed) concentrated on the advantaged groups (while noting, at p.250, that the Court of Appeal had considered both groups).

The identification of the pool

    56.  In order to determine this appeal your Lordships do have to look at the question which was left open in Seymour-Smith. There are as I see it two issues. One is the identification of the correct pool. The other is the choice (or balance) between what I have called the advantage-led approach and the disadvantage-led approach. The identification of the pool seems to have dominated the arguments below, but in my opinion it can now be seen as relatively uncontroversial and relatively straightforward. I shall therefore try to dispose of it at once.

    57.  The relevant legislation applies generally to all employed persons on whom rights are conferred by the Employment Rights Act 1996. As a matter of general principle, therefore, the pool should be all those persons. There was argument below as to whether the pool should include (i) persons who had not yet completed their qualifying period of service and/or (ii) persons who were not employed but were seeking employment. But any such fine-tuning would have only the most marginal effect on the figures. That is apparent from two tables attached to the further witness statement of Dr Hotopp, the Secretary of State's expert statistician. The differences between the two sets of figures are insignificant (in the general, non-technical sense: as to the dangers of lawyers misunderstanding statistical significance see Douglas W. Vick, Statistical Significance and the Significance of Statistics (2000) 116 LQR 575).

    58.  The controversy over the identification of the pool seems to have originated partly from Lindsay J's understandable wish (in "Rutherford 1" [2002] ICR 123) to give guidance to the Employment Tribunal to which the case was to be remitted. In his summary (para 18) he encouraged the use of more than one type of comparison, if only as a check. I see nothing obviously wrong with that general guidance, so long as it does not lead to over-elaboration. But in this case it unfortunately produced (in para 16 of the Employment Tribunal's decision) a long and (if I may say so) rather breathless and incoherent discussion about the appropriate methodology. Mr Allen QC (for the appellants) helpfully suggested that it should be read as four different paragraphs discussing different approaches to the problem. Read in that way, it becomes clear that the notion of "those for whom retirement by 65 has some real meaning" (language criticised by Mr Pannick QC for the Secretary of State as introducing "a sociological concept of considerable subjectivity and uncertainty") was only one of several possible approaches, and I respectfully think that the Court of Appeal was wrong to suppose (see para 19 of the judgment of Mummery LJ) that the Employment Tribunal had identified the pool as employed persons aged between 55 and 74. But in any event this is something of an irrelevance, since starting the pool with 55-year-olds makes the figures (slightly) less helpful for the appellants. The reason for that is apparent from the first three lines of each of the subsidiary tables in Table A1 annexed to the Employment Tribunal's decision: by age 55 large numbers of women have already left the labour market, and men outnumber them by about 9 to 8.

Advantage-led or disadvantage-led?

    59.  Later in the discussion in paragraph 16 of the Employment Tribunal's decision there are two tables (for 1998 and 2001, the years in which Mr Rutherford and Mr Bentley respectively were dismissed). These tables seem to have carried a good deal of weight in the Employment Tribunal's final conclusion. They do treat the whole labour force (between the ages of 16 and 79) as the relevant pool. I reproduce that for 1998 (with a small addition clarifying the "gender ratio" column); that for 2001 is very similar. The decision drew particular attention to the final gender ratio (1.44: 1 for Mr Rutherford;

    the corresponding figure for Mr Bentley was 1.53: 1).

  Males Females Gender Ratio
13,588,800 12,546,400 1.08 : 1
195,200 124,900 1.56 : 1
Disadvantaged as
% of advantaged
1.44% 1.00% 1.44 : 1

    60.  I have two comments on this table. One is that it is usual to regard the pool as consisting of the aggregate of the advantaged and the disadvantaged, and to express the proportions of the advantaged and the disadvantaged as percentages of the whole pool: then (as in the tables in the speeches of Lord Slynn and Lord Nicholls in Seymour-Smith) the percentages will always total 100, which is easier to understand. In the above table that modification makes almost no difference to the figures for the disadvantaged, because almost all the pool, men and women, are advantaged: the figures become men 1.42%, women 0.99% (with the ratio unchanged). The other (and more important) comment is that the table omits the corresponding figures for the advantaged, which (as proportions of the whole pool) are men 98.58%, women 99.01%, with a ratio (M/F) of 1:1.004. So when related to the pool as a whole, the advantage to women is barely perceptible. The revised table below shows the effect of making these changes.

  Males Females Gender Ratio
Advantaged 13,588,800 12,546,400 1.08 : 1
Disadvantaged 195,200 124,900 1.56 : 1
Total pool 13,784,000 12,671,300 1.08 : 1
Disadvantaged as
% of pool
1.42% 0.99% 1.44 : 1
Advantaged as %
of pool
98.58% 99.01% 1 : 1.004

    61.  This point was (as Lord Slynn noted when Seymour-Smith was before this House) discussed by the Divisional Court and the Court of Appeal in Seymour-Smith, but only as a curiosity, since the difficulties in that case were of a different character. The clearest comment was that of McCullough J in the Divisional Court [1995] ICR 889, 914 (though he was giving his view without the benefit of the guidance later obtained from the Court of Justice):

    "My conclusion is that, in considering whether there is considerable disparity, the Court should look both at the relative percentages of those who meet the requirement and at the relative percentages of those who do not. Of these the most important group will be those who qualify. The following example makes the point. If 98% of men qualify and 2% do not, and if 96% of women qualify and 4% do not, it would not be right to conclude that the disparity was considerable. But if only 4% of men and only 2% of women qualified the opposite conclusion might well be correct."

    62.  It may be helpful to some if I summarise this in algebraic form. If A and D denote the numbers of advantaged and disadvantaged persons, and M and W the numbers of men and women, the men's pool (PM) consists of AM + DM and the women's pool of AW + DW. The proportions to be considered can then be expressed as the percentages AM/PM % ("advantage-led") or DM/PM % ("disadvantage-led"); similarly with the women's pool. Mummery LJ has in para 25 of his judgment given a clear and detailed description of this advantage-led approach. In many cases (including Seymour-Smith) either approach will produce much the same result. But when the vast majority of each pool is classified as advantaged, the choice (or balance) between the two methods becomes critical. 1.44:1 is arguably a considerable difference, but 1: 1.004 is not.

    63.  My Lords, the jurisprudence of the Court of Justice does not provide a clear answer to this question. The formulation in para 60 of its judgment in Seymour-Smith has been repeated in numerous later cases, including cases decided since the coming into full force of the Burden of Proof Directive. It is sufficient to mention the decision of the Grand Chamber in Allonby v Accrington and Rossendale College [2004] ICR 1328, 1361, para 79 (that case was decided on facts antedating the Directive, but it seems most unlikely that the Grand Chamber would not have referred to it if it had been thought to have made a significant change in the law). But (as already noted) none of these cases involved pools with a very large percentage of advantaged persons and a very small percentage of disadvantaged persons (or vice versa).

    64.  Mr Allen placed great reliance on the Burden of Proof Directive, and submitted that in the face of the plain language of article 2 it would be "contra legem" to focus on a comparison of the advantaged percentages in the two groups. But the Burden of Proof Directive was intended to codify, not to alter, the jurisprudence of the Court of Justice. Article 2 defines indirect discrimination in terms of disadvantage, which is unsurprising, but (as Mummery LJ observed in para 35 of his judgment in the Court of Appeal),

    "The definition describes when a certain state of affairs (ie indirect discrimination) exists: it does not, however, prescribe the methodology for assessing the statistical evidence in order to determine whether or not that state of affairs exists. No methodology has been laid down in the Treaty or in any Directive or in national legislation. It has been left to the national courts and tribunals, which hear and assess the evidence and find the facts, to work out from case to case a satisfactory method for assessing whether or not there is disparate adverse impact in the particular case."

    65.  Your Lordships must therefore resolve this matter largely as a matter of principle. I must confess that my first reaction was that in this case an advantage-led approach could have the effect of masking a significant degree of disparate disadvantage, and that a demonstrable element of indirect discrimination should not be ignored merely because it affected relatively few people. Plainly that is the case with direct discrimination, whether on the ground of gender or on any other ground. To say that only a few individuals are disadvantaged is no excuse. But indirect discrimination is in my view different, especially where it is the type which may be produced by national legislation of general application, and which may be assessed on a purely statistical basis: see Allonby [2004] ICR 1328, 1362, para 80. In such cases the pool should be taken as it is, and the relevant statistics analysed, without regard to any underlying sociological or economic factors. Such factors may be highly relevant at the second stage of objective justification, but they are not relevant at the first stage. For that reason the Employment Tribunal was in my view wrong to ask itself (at the first stage) who had a real interest in retirement by 65 (though that point by itself made little practical difference to the outcome). The purpose of the analysis is to look for disproportionate impact on men as such and on women as such, as anonymous, characterless material for statisticians.

    66.  Our domestic jurisprudence (concerned as it is with anti-discrimination legislation framed in much more detailed terms than EU law) has gone rather further than the Court of Justice in exploring the intricacies of indirect discrimination. It has recognised that the appropriate pool should not be narrowed by reference to qualifications or conditions which are not logically relevant to the statistical exercise in hand : see Perera v Civil Service Commission (No 2) [1982] ICR 350, 358-359 (proportion of late immigrants irrelevant; upheld by the Court of Appeal [1983] ICR 428 without mention of this point); University of Manchester v Jones [1993] ICR 474, 495 A-C and 505 E-H (proportion of "mature" graduates irrelevant); London Underground v Edwards (No 2) [1999] ICR 494, 508H-509A (proportion of single parents irrelevant); and Barry v Midland Bank plc in the Court of Appeal [1999] ICR 319, 334B and in this House [1999] ICR 859, 869H - 870A (pool incorrectly limited to one specialised department). This is certainly consistent with, and to my mind provides a little support for, an advantage-led approach which takes the unvarnished statistics as it finds them.

    67.  I do not express the view that some element of disadvantage-led analysis may not be appropriate in some cases. But it must be recognised that there is a difficulty here: the more extreme the majority of the advantaged in both pools, the more difficult it is, with any intellectual consistency, to pay much attention to the result of a disadvantage-led approach. However I can imagine some (perhaps improbable) cases in which a disadvantage-led approach would serve as an alert to the likelihood of objectionable discrimination. If (in a pool of one thousand persons) the advantaged 95% were split equally between men and women, but the disadvantaged 5% were all women, the very strong disparity of disadvantage would, I think, make it a special case, and the fact that the percentages of the advantaged were not greatly different (100% men and 90.5% women) would not be decisive.

    68.  But that is not this case. A proportion of disadvantage of 1.44: 1 is not such an extraordinary disparity as to make it necessary to resile from the absence of any significant disparity (1: 1.004) in the two advantaged groups. The Employment Tribunal erred in law in ignoring completely the effect of an advantage-led approach.

    69.  For these reasons (which are largely but not entirely the same as those of the Court of Appeal) I would dismiss this appeal. Although your Lordships had full written and oral submissions on the issue of objective justification, I think it better not to express any views on that issue.

    70.  I prepared this opinion without having had the advantage of considering in draft the opinions of my noble and learned friends Lord Scott of Foscote, Lord Rodger of Earlsferry and Baroness Hale of Richmond. They would dispose of this appeal on broader grounds and so avoid entanglement in arguments which many may regard as over-elaborate, artificial and inappropriate for a subject of such social and economic importance as discrimination in employment. I would very much prefer to find a simpler and more transparent principle on which to decide the appeal, and by which tribunals can be guided in future. But the grounds on which my noble and learned friends would decide the appeal were not argued below or in this House. Nor do I find it easy to extract from their opinions a single easily-stated principle.


    71.  The essence of indirect discrimination is that an apparently neutral requirement or condition (under the old formulation) or provision, criterion or practice (under the new) in reality has a disproportionate adverse impact upon a particular group. It looks beyond the formal equality achieved by the prohibition of direct discrimination towards the more substantive equality of results. A smaller proportion of one group can comply with the requirement, condition or criterion or a larger proportion of them are adversely affected by the rule or practice. This is meant to be a simple objective enquiry. Once disproportionate adverse impact is demonstrated by the figures, the question is whether the rule or requirement can objectively be justified.

    72.  It is of the nature of such apparently neutral criteria or rules that they apply to everyone, both the advantaged and the disadvantaged groups. So it is no answer to say that the rule applies equally to men and women, or to each racial or ethnic or national group, as the case may be. The question is whether it puts one group at a comparative disadvantage to the other. However, the fact that more women than men, or more whites than blacks, are affected by it is not enough. Suppose, for example, a rule requiring that trainee hairdressers be at least 25 years old. The fact that more women than men want to be hairdressers would not make such a rule discriminatory. It would have to be shown that the impact of such a rule worked to the comparative disadvantage of would-be female or male hairdressers as the case might be.

    73.  But the notion of comparative disadvantage or advantage is not straightforward. It involves defining the right groups for comparison. The twists and turns of the domestic case law on indirect discrimination show that this is no easy matter. But some points stand out. First, the concept is normally applied to a rule or requirement which selects people for a particular advantage or disadvantage. Second, the rule or requirement is applied to a group of people who want something. The disparate impact complained of is that they cannot have what they want because of the rule or requirement, whereas others can.

    74.  What is the comparative advantage and disadvantage in this case? It cannot simply be being under or over the age of 65. That in itself is neither an advantage nor a disadvantage, until it is linked to what the people concerned want to have or not to have. If one wants to have a pension, then reaching pensionable age is an advantage. If one wants to go on working beyond pensionable age, then reaching that age may be a disadvantage.

    75.  The advantage or disadvantage in question here is going on working over the age of 65 while still enjoying the protection from unfair dismissal and redundancy that younger employees enjoy. As Mr Allen QC for the appellants pointed out, that protection has an impact, not only when employment comes to an end, but also upon whether or not it is brought to an end, and if so, how.

    76.  If that is so, it matters not that there are other men and women who have left the workforce at an earlier age and are thus uninterested in whether or not they will continue to be protected. The people who want the protection are the people who are still in the workforce at the age of 65. And the rule has no disproportionate effect upon any particular group within that group. It applies to the same proportion of women in that group as it applies to men. There is no comparison group who wants this particular benefit and can more easily obtain it.

    77.  The appellants cannot object that this approach defines the advantage and disadvantage by reference to the very rule which is under attack; on the contrary, that is exactly what they have sought to do by treating the advantage and disadvantage as being under or over the age of 65. But the result is the same even if one defines the advantage and disadvantage without reference to the age limit. The advantage is to be able to enjoy protection against unfair dismissal and redundancy throughout one's working life. As my noble and learned friend Lord Walker of Gestingthorpe has demonstrated, the sex differential between those who can and those who cannot do this is negligible, even though there are more men in the group who cannot. But in my view one should not be bringing into the comparison people who have no interest in the advantage in question.

    78.  This approach, defining advantage and disadvantage by reference to what people want, chimes with the definition of discrimination given by McIntyre J in the seminal Canadian case of Andrews v British Columbia [1989] 1 SCR 143:

    ". . . discrimination may be described as a distinction, whether intentional or not but based on grounds relating to the personal characteristics of the individual or group, which has the effect of imposing burdens, obligations or disadvantages on such individual or group not imposed upon others, or which withholds or limits access to opportunities, benefits and advantages available to other members of society." (Emphasis supplied)

It also chimes with Sandra Fredman's observation, in Discrimination Law (Clarendon Law Series, 2002, p 115), that "A disparate impact is not itself discriminatory. Unequal results are legitimate if no exclusionary barrier can be identified . . ." The sorts of cases where indirect discrimination can be established confirm this.

    79.  Say members of both groups want a particular job, but age limits, educational qualifications, or size, height or physical strength requirements are applied (as in the progenitor of the indirect discrimination concept, Griggs v Duke Power 401 US 424). These may well have a disproportionate adverse impact upon one sex, or upon different racial, ethnic or national groups. The question then is whether they can be independently justified.

    80.  Say members of both groups want compensation for the loss of their jobs, but length of employment criteria, or "last in first out" practices, or preference for full-timers over part-timers, apply. These too may well have a disproportionate adverse impact upon one sex, or upon different racial, ethnic or national groups. Again the question is whether they can be independently justified.

    81.  Say members of both groups want various work-related benefits, such as occupational pensions, sick pay, holiday pay, bonuses of one sort or another, but there are length of employment criteria or differential treatment between full-timers and part-timers. These again may have a disproportionate adverse impact upon one sex, or upon different racial ethnic or national groups. Again the question is whether they can be independently justified.

    82.  The common feature is that all these people are in the pool who want the benefit - or not to suffer the disadvantage - and they are differentially affected by a criterion applicable to that benefit or disadvantage. Indirect discrimination cannot be shown by bringing into the equation people who have no interest in the advantage or disadvantage in question. If it were, one might well wish to ask whether the fact that they were not interested was itself the product of direct or indirect discrimination in the past.

    83.  For these reasons, which are essentially the same as those given by my noble and learned friends, Lord Scott of Foscote and Lord Rodger of Earlsferry, I too would dismiss these appeals without requiring the Department to justify the rule.


Lords  Parliament  Commons  Search  Contact Us  Index 

© Parliamentary copyright 2006
Prepared 4 May 2006